deliverable/binutils-gdb.git
4 years agogdb: protect some 'regcache_read_pc' calls
Tankut Baris Aktemur [Thu, 14 May 2020 11:59:53 +0000 (13:59 +0200)] 
gdb: protect some 'regcache_read_pc' calls

It possible that a thread whose PC we attempt to read is already dead.
In this case, 'regcache_read_pc' errors out.  This impacts the
"proceed" execution flow, where GDB quits early before having a chance
to check if there exists a pending event.  To remedy, keep going with
a 0 value for the PC if 'regcache_read_pc' fails.  Because the value
of PC before resuming a thread is mostly used for storing and checking
the next time the thread stops, this tolerance is expected to be
harmless for a dead thread/process.

gdb/ChangeLog:
2020-05-14  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

* regcache.c (regcache_read_pc_protected): New function
implementation that returns 0 if the PC cannot read via
'regcache_read_pc'.
* infrun.c (proceed): Call 'regcache_read_pc_protected'
instead of 'regcache_read_pc'.
(keep_going_pass_signal): Ditto.

gdbsupport/ChangeLog:
2020-05-14  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

* common-regcache.h (regcache_read_pc_protected): New function
declaration.

4 years agoUpdate Swedish translation for the gas sub-directory and a new Serbian translation...
Nick Clifton [Thu, 14 May 2020 10:26:26 +0000 (11:26 +0100)] 
Update Swedish translation for the gas sub-directory and a new Serbian translation for the gold sub-directory.

4 years agoRISC-V: Add elfNN_riscv_mkobject to initialize RISC-V tdata.
Nelson Chu [Wed, 13 May 2020 08:05:45 +0000 (16:05 +0800)] 
RISC-V: Add elfNN_riscv_mkobject to initialize RISC-V tdata.

For now we only have one char pointer in RISC-V tdata, so it should be fine.
But once we need more elements in tdata, then we may get some uninitialize
or unexpected values.  I do meet the same problem when extending the RISC-V
tdata.

bfd/
elfnn-riscv.c (elfNN_riscv_mkobject):  New function.  We need this
to initialize RISC-V tdata.

4 years agoAutomatic date update in version.in
GDB Administrator [Thu, 14 May 2020 00:00:16 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years agoRemove ada-lang.c:align_value
Tom Tromey [Wed, 13 May 2020 19:15:13 +0000 (13:15 -0600)] 
Remove ada-lang.c:align_value

I recently noticed the align_value function in ada-lang.c.  This can
be removed, in favor of align_up from gdbsupport.

gdb/ChangeLog
2020-05-13  Tom Tromey  <tromey@adacore.com>

* ada-lang.c (align_value): Remove.
(ada_template_to_fixed_record_type_1): Use align_up.

4 years agogdb: update the copyright year in async-event.[ch]
Tankut Baris Aktemur [Wed, 13 May 2020 12:53:37 +0000 (14:53 +0200)] 
gdb: update the copyright year in async-event.[ch]

The async-event.[ch] files were introduced recently as a result of
splitting the event-loop.  I believe the copyright year update was
just an oversight.  So, this patch fixes that.

gdb/ChangeLog:
2020-05-13  Tankut Baris Aktemur  <tankut.baris.aktemur@intel.com>

* async-event.c: Update the copyright year.
* async-event.h: Update the copyright year.

4 years agoHave the linker's help text include the default setting of the --hash-style option...
Nick Clifton [Wed, 13 May 2020 09:59:02 +0000 (10:59 +0100)] 
Have the linker's help text include the default setting of the --hash-style option, if relevent.

PR 25979
* lexsup.c (elf_shlib_list_options): Include the default value for
the hash style in the output text.

4 years agoSync config and libiberty with GCC
H.J. Lu [Wed, 13 May 2020 01:37:03 +0000 (18:37 -0700)] 
Sync config and libiberty with GCC

config/

PR bootstrap/94998
* cet.m4 (GCC_CET_HOST_FLAGS): Enable CET in cross compiler if
possible.

libiberty/

PR bootstrap/94998
* configure: Regenerated.

4 years agoAutomatic date update in version.in
GDB Administrator [Wed, 13 May 2020 00:00:18 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years agogdb/testsuite: Disable path and duplicate checks when parallel testing
Andrew Burgess [Tue, 12 May 2020 16:38:17 +0000 (17:38 +0100)] 
gdb/testsuite: Disable path and duplicate checks when parallel testing

This commit disables the recently added checking for paths in test
names, and for duplicate test names, when the gdb tests are run in
parallel.

When running the gdb tests in parallel the extra result count lines
produced cause the dg-extract-results scripts to exit with an error.

The patches for the dg-extract-results scripts have been posted to the
gcc-patches mailing list here:

https://gcc.gnu.org/pipermail/gcc-patches/2020-May/545562.html

Once they are merged there then these changes can be merged over to
binutils-gdb, and this commit can be reverted.

gdb/testsuite/ChangeLog:

* lib/check-test-names.exp: Disable when testing is being run in
parallel.

4 years agogdb: make two objfile functions return bool
Simon Marchi [Tue, 12 May 2020 15:17:01 +0000 (11:17 -0400)] 
gdb: make two objfile functions return bool

gdb/ChangeLog:

* objfiles.h (is_addr_in_objfile,
shared_objfile_contains_address_p): Return bool.
* objfile.c (is_addr_in_objfile,
shared_objfile_contains_address_p): Return bool.

4 years ago[gdb/testsuite] Fix incorrect string concat in jit-elf.exp
Tom de Vries [Tue, 12 May 2020 12:59:01 +0000 (14:59 +0200)] 
[gdb/testsuite] Fix incorrect string concat in jit-elf.exp

When running test-case gdb.base/jit-elf.exp with target board
cc-with-gdb-index, we get:
...
spawn -ignore SIGHUP gdb/contrib/cc-with-tweaks.sh -i gcc \
  -fno-stack-protector src/gdb/testsuite/gdb.base/jit-elf-main.c \
  -fdiagnostics-color=never -DATTACH=1 -DLOAD_ADDRESS=0x7000000 \
  -DLOAD_INCREMENT=0x1000000 -g -lm \
  -o outputs/gdb.base/jit-elf/jit-elf-main"-attach"^M
outputs/gdb.base/jit-elf/.tmp/jit-elf-main-attach: \
  No such file or directory.^M
output is:
outputs/gdb.base/jit-elf/.tmp/jit-elf-main-attach: \
  No such file or directory.^M

gdb compile failed, outputs/gdb.base/jit-elf/.tmp/jit-elf-main-attach: \
  No such file or directory.
UNTESTED: gdb.base/jit-elf.exp: failed to compile jit-elf-main"-attach"
...

The problem is a string concat in jit-elf.exp:
...
  ${main_binfile}"-attach"
...
which is intended to generate string 'jit-elf-main-attach' but instead
generates string 'jit-elf-main"-attach"'.

Fix this by using "${main_binfile}-attach" instead.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* gdb.base/jit-elf.exp: Fix string concat.

4 years ago[gdb/testsuite] Fix tcl error in jit-elf-helpers.exp
Tom de Vries [Tue, 12 May 2020 12:41:47 +0000 (14:41 +0200)] 
[gdb/testsuite] Fix tcl error in jit-elf-helpers.exp

When running test-case jit-elf.exp with target board cc-with-gdb-index, I run
into:
...
gdb compile failed, outputs/gdb.base/jit-elf/.tmp/jit-elf-main-attach: \
  No such file or directory.
ERROR: tcl error sourcing src/gdb/testsuite/gdb.base/jit-elf.exp.
ERROR: can't read "main_basename": no such variable
    while executing
"untested "failed to compile ${main_basename}.c""
    (procedure "compile_jit_main" line 7)
...

The problem is in compile_jit_main in lib/jit-elf-helpers.exp, where we try to
emit an untested message using global variable main_basename.c.

Fixing this by declaring the variable global results in duplicate test-names,
because the same source file is compiled more than once.

Instead, fix this by using the result name in the untested message.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* lib/jit-elf-helpers.exp: Don't use undefined variables in untested
messages.

4 years ago [PATCH] bfd: Fix 64-bit relocation handling for a.out
Gunther Nikl [Tue, 12 May 2020 11:18:09 +0000 (12:18 +0100)] 
 [PATCH] bfd: Fix 64-bit relocation handling for a.out

* aoutx.h (NAME (aout, swap_std_reloc_out)): Reject an unsupported
relocation size.

4 years ago[gdb/testsuite] Fix duplicate test-names in gdb.multi
Tom de Vries [Tue, 12 May 2020 09:18:14 +0000 (11:18 +0200)] 
[gdb/testsuite] Fix duplicate test-names in gdb.multi

In gdb.multi we have:
...
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=run: inf2_how=run: info inferiors
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=run: inf2_how=attach: info inferiors
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=run: inf2_how=tty: info inferiors
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=attach: inf2_how=run: info inferiors
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=attach: inf2_how=attach: attach
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=attach: inf2_how=attach: info inferiors
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=attach: inf2_how=tty: info inferiors
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=tty: inf2_how=run: info inferiors
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=tty: inf2_how=attach: info inferiors
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=tty: inf2_how=tty: tty TTY
DUPLICATE: gdb.multi/multi-term-settings.exp: \
  inf1_how=tty: inf2_how=tty: info inferiors
...

Fix these using with_test_prefix.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* gdb.multi/multi-term-settings.exp: Use with_test_prefix.

4 years ago[gdb/testsuite] Fix duplicate test-names in gdb.ada
Tom de Vries [Tue, 12 May 2020 09:05:44 +0000 (11:05 +0200)] 
[gdb/testsuite] Fix duplicate test-names in gdb.ada

In gdb.ada we have these duplicate test-names:
...
DUPLICATE: gdb.ada/catch_ex.exp: continuing to program completion
DUPLICATE: gdb.ada/mi_catch_ex.exp: breakpoint at main
DUPLICATE: gdb.ada/mi_catch_ex.exp: mi runto main
DUPLICATE: gdb.ada/mi_catch_ex_hand.exp: breakpoint at main
DUPLICATE: gdb.ada/mi_catch_ex_hand.exp: mi runto main
...

Fix these using with_test_prefix.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* gdb.ada/catch_ex.exp: Use with_test_prefix.
* gdb.ada/mi_catch_ex.exp: Same.
* gdb.ada/mi_catch_ex_hand.exp: Same.

4 years ago[gdb/testsuite] Fix duplicate test-names in gdb.fortran
Tom de Vries [Tue, 12 May 2020 08:53:47 +0000 (10:53 +0200)] 
[gdb/testsuite] Fix duplicate test-names in gdb.fortran

In gdb.fortran we have:
...
DUPLICATE: gdb.fortran/complex.exp: whatis $
DUPLICATE: gdb.fortran/complex.exp: whatis $
DUPLICATE: gdb.fortran/complex.exp: whatis $
DUPLICATE: gdb.fortran/complex.exp: whatis $
...

Fix this by using with_test_prefix.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* gdb.fortran/complex.exp: Use with_test_prefix.

4 years ago[gdb/testsuite] Fix duplicate test-names in gdb.trace
Tom de Vries [Tue, 12 May 2020 08:45:20 +0000 (10:45 +0200)] 
[gdb/testsuite] Fix duplicate test-names in gdb.trace

In gdb.trace we have these duplicates:
...
DUPLICATE: gdb.trace/passcount.exp: 4.20a: set all three passcounts to three
DUPLICATE: gdb.trace/passcount.exp: 4.6: set passcount to zero
DUPLICATE: gdb.trace/passcount.exp: 4.7: set passcount to large number (32767)
...

Fix these by fixing the test-names.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* gdb.trace/passcount.exp: Fix test-names.

4 years ago[gdb/testsuite] Fix duplicate test-names in gdb.pascal
Tom de Vries [Tue, 12 May 2020 08:37:04 +0000 (10:37 +0200)] 
[gdb/testsuite] Fix duplicate test-names in gdb.pascal

In gdb.pascal we have these duplicates:
...
DUPLICATE: gdb.pascal/gdb11492.exp: next
DUPLICATE: gdb.pascal/gdb11492.exp: print char_array
...

Fix these by using with_test_prefix.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* gdb.pascal/gdb11492.exp: Use with_test_prefix.

4 years ago[gdb/testsuite] Fix duplicate test-names in gdb.{gdb,opt,xml}
Tom de Vries [Tue, 12 May 2020 08:28:44 +0000 (10:28 +0200)] 
[gdb/testsuite] Fix duplicate test-names in gdb.{gdb,opt,xml}

There are 3 test directories with one duplicate test-name: gdb.gdb, gdb.opt
and gdb.xml.  The duplicates are:
...
DUPLICATE: gdb.gdb/complaints.exp: call complaint_internal ($cstr)
DUPLICATE: gdb.opt/inline-locals.exp: info locals above bar 2 \
  (PRMS: gdb/25695)
DUPLICATE: gdb.xml/tdesc-regs.exp: ptype $extrareg
...

Fix as appropriate.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* gdb.gdb/complaints.exp: Use with_test_prefix.
* gdb.xml/tdesc-regs.exp: Same.
* gdb.opt/inline-locals.exp: Fix test name.

4 years ago[gdb/testsuite] add jit-elf-util.h and run jit function
Mihails Strasuns [Tue, 18 Feb 2020 11:30:54 +0000 (12:30 +0100)] 
[gdb/testsuite] add jit-elf-util.h and run jit function

Splits ELF related symbols into a separate jit-elf-util.h header and
enhances it with a few more.

Intention is to make adding new JIT tests possible without repeating
most of the common boilerplate.

As a test enhancement, jit-elf-main.c now calls the renamed function
after registering the jit object and ensures it returns an expected
result.

gdb/testsuite/ChangeLog:

2020-02-18  Mihails Strasuns  <mihails.strasuns@intel.com>

* gdb.base/jit-elf-util.h: New header file.
* gdb.base/jit-elf-main.c: Use jit-elf-util.h, add a call to
  the renamed JIT function to verify its result.

4 years ago[gdb/testsuite] define jit function name via macro
Mihails Strasuns [Fri, 27 Mar 2020 10:21:01 +0000 (11:21 +0100)] 
[gdb/testsuite] define jit function name via macro

Replaces previous approach with patching resulting ELF binary after
loading - now that each test iteration works on a separately compiled
binary it is not necessary anymore.

Tests are still being ran without debug info to preserve original test
functionality but this change opens up the possibility to enable debug
info if needed too.

gdb/testsuite/ChangeLog:

2020-03-27  Mihails Strasuns  <mihails.strasuns@intel.com>

* lib/jit-elf-helpers.exp: Supply -DFUNCTION_NAME macro
  definition when compiling jit-elf-solib.co.
* gdb.base/jit-elf-main.c: Stop patching jit function name.
* gdb.base/jit-elf-solib.c: Use FUNCTION_NAME macro value as a
  function name.

4 years ago[gdb/testsuite] use -Ttext-segment for jit-elf tests
Mihails Strasuns [Mon, 17 Feb 2020 10:36:47 +0000 (11:36 +0100)] 
[gdb/testsuite] use -Ttext-segment for jit-elf tests

Removes the need to manually relocate loaded ELF binary by using a fixed
constant as both mmap base address and as a requested first segment
address supplied to the linker.

In future will enable JIT tests with a valid DWARF debug info.  Current
tests still need to compile without a debug info though, because they do
a function name modification.

gdb/testsuite/ChangeLog:

2020-02-18  Mihails Strasuns  <mihails.strasuns@intel.com>

* lib/jit-elf-helpers.exp: Supply -Ttext-segment linker flag and
  define LOAD_ADDRESS/LOAD_INCREMENT macros for the compiled binaries.
* gdb.base/jit-elf-main.c: Use LOAD_ADDRESS/LOAD_INCREMENT to
  calculate the mmap address.

4 years ago[gdb/testsuite] add lib/jit-elf-helpers.exp
Mihails Strasuns [Mon, 17 Feb 2020 10:24:14 +0000 (11:24 +0100)] 
[gdb/testsuite] add lib/jit-elf-helpers.exp

New utility library to be used by jit-elf tests responsible for
compiling binary artifacts. In the next commit the compilation process
will become more complicated because of extra mandatory flag - keeping
it in one place will make tests less fragile.

gdb/testsuite/ChangeLog:

2020-02-18  Mihails Strasuns  <mihails.strasuns@intel.com>

* lib/jit-elf-helpers.exp: New file.
* gdb.base/jit-elf.exp: Updated to use jit-elf-helpers.exp.
* gdb.base/jit-elf-so.exp: Updated to use jit-elf-helpers.exp.

4 years ago[gdb/testsuite] use args as lib list for jit-elf tests
Mihails Strasuns via Gdb-patches [Tue, 31 Mar 2020 17:13:51 +0000 (19:13 +0200)] 
[gdb/testsuite] use args as lib list for jit-elf tests

Old usage: jit-elf-main lib.so 2
New usage: jit-elf-main lib.so.1 lib.so.2

Refactoring necessary to support running tests over multiple jit
binaries rather than mapping the same binary muultiple times.

gdb/testsuite/ChangeLog:

2020-02-18  Mihails Strasuns  <mihails.strasuns@intel.com>

* gdb.base/jit-elf-main.c: Read lib list from argc/argv.
* gdb.base/jit-elf.exp: Compile N jit libraries and use the list.
* gdb.base/jit-elf-so.exp: Ditto.

4 years ago[gdb/testsuite] Fix duplicate test-names in gdb.dwarf2
Tom de Vries [Tue, 12 May 2020 07:47:15 +0000 (09:47 +0200)] 
[gdb/testsuite] Fix duplicate test-names in gdb.dwarf2

We currently have these duplicate test-names in gdb.dwarf2:
...
$ grep ^DUPLICATE: gdb.sum
DUPLICATE: gdb.dwarf2/comp-unit-lang.exp: show language
DUPLICATE: gdb.dwarf2/data-loc.exp: get integer valueof "sizeof (int)"
DUPLICATE: gdb.dwarf2/data-loc.exp: get integer valueof "sizeof (void *)"
DUPLICATE: gdb.dwarf2/data-loc.exp: get integer valueof "sizeof (int)"
DUPLICATE: gdb.dwarf2/data-loc.exp: get integer valueof "sizeof (int)"
DUPLICATE: gdb.dwarf2/data-loc.exp: ptype foo.array_type
DUPLICATE: gdb.dwarf2/varval.exp: get integer valueof "sizeof (int)"
DUPLICATE: gdb.dwarf2/varval.exp: get integer valueof "sizeof (void *)"
DUPLICATE: gdb.dwarf2/implref-struct.exp: print-object=off: \
  set print object off
DUPLICATE: gdb.dwarf2/implref-struct.exp: print-object=on: \
  set print object on
DUPLICATE: gdb.dwarf2/dw2-bad-parameter-type.exp: ptype f
...

Fix as appropriate.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-12  Tom de Vries  <tdevries@suse.de>

* gdb.dwarf2/comp-unit-lang.exp: Use with_test_prefix.
* gdb.dwarf2/dw2-bad-parameter-type.exp: Same.
* gdb.dwarf2/implref-struct.exp: Same.
* gdb.dwarf2/varval.exp: Ensure get_sizeof is called once per type.
* gdb.dwarf2/data-loc.exp: Same.  Remove duplicate test.

4 years agogdb: change duplicate test name in gdb.base/jit-so.exp
Simon Marchi [Tue, 12 May 2020 00:10:35 +0000 (20:10 -0400)] 
gdb: change duplicate test name in gdb.base/jit-so.exp

When running gdb.base/jit-so.exp, we see duplicate test names:

$ grep PASS testsuite/gdb.sum | sort | uniq -c | sort -n
      ...
      2 PASS: gdb.base/jit-so.exp: one_jit_test-1: info function jit_function
      2 PASS: gdb.base/jit-so.exp: one_jit_test-2: info function jit_function

Give an explicit name to one test to avoid this.

gdb/testsuite/ChangeLog:

* gdb.base/jit-so.exp (one_jit_test): Change test name.

4 years agoAutomatic date update in version.in
GDB Administrator [Tue, 12 May 2020 00:00:11 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years agoRegen ld/Makefile.in
Alan Modra [Mon, 11 May 2020 23:48:06 +0000 (09:18 +0930)] 
Regen ld/Makefile.in

Missed from c578f16ef18fd.

4 years agoRestore info_command and breakpoint
Tom Tromey [Mon, 11 May 2020 21:28:12 +0000 (15:28 -0600)] 
Restore info_command and breakpoint

As discussed on gdb-patches, this restores info_command and the
breakpoint on info_command in gdb-gdb.gdb.  This reverts a tiny part
of 0743fc83c03 ("Replace most calls to help_list and cmd_show_list"),
as well as 652fc23a30a ("Remove gdb-gdb.gdb breakpoint on disappeared
function info_command.").

gdb/ChangeLog
2020-05-11  Tom Tromey  <tromey@adacore.com>

* cli/cli-cmds.c (info_command): Restore.
(_initialize_cli_cmds): Use add_prefix_command for "info".
* gdb-gdb.gdb.in: Restore breakpoint on info_command.

4 years agogdb/testsuite: Detect and warn about duplicate test names
Andrew Burgess [Thu, 23 Apr 2020 14:19:15 +0000 (15:19 +0100)] 
gdb/testsuite: Detect and warn about duplicate test names

Building on the previous commit, this patch detects when two tests
have the same test name and causes Dejagnu to print a new result type
'# of duplicate test names' in the result summary.  A line starting
with 'DUPLICATE: ' is also added to the gdb.sum and gdb.log files.

The DUPLICATE markers will be printed the second time a duplicate test
name is seen, and every time after that.  So you might see:

  PASS: gdb.base/sometest.exp: foo
  PASS: gdb.base/sometest.exp: bar
  PASS: gdb.base/sometest.exp: foo
  DUPLICATE: gdb.base/sometest.exp: foo
  PASS: gdb.base/sometest.exp: baz
  PASS: gdb.base/sometest.exp: foo
  DUPLICATE: gdb.base/sometest.exp: foo

However, the results will report a duplicate count of 1, indicating
that just one test name (foo) was duplicated.

Currently if the tests are run in parallel mode the new result type is
not merged into the combined summary file so users will need to run in
non-parallel mode to check this result.  Similarly, the 'DUPLICATE: '
markers will not be merged into the final gdb.sum file.  A later
commit will fix this.

gdb/testsuite/ChangeLog:

* lib/check-test-names.exp (all_test_names): New module variable.
(counts): Add 'duplicates' field.
(_check_duplicates): New procedure.
(check): Also check for duplicates.
(do_log_summary): Print duplicates count.
(do_reset_vars): Reset counter for duplicate test names, and
discard all know test names.

4 years agogdb/testsuite: Detect and warn if paths are used in test names
Andrew Burgess [Thu, 23 Apr 2020 09:47:29 +0000 (10:47 +0100)] 
gdb/testsuite: Detect and warn if paths are used in test names

A new library is introduced that hooks into the core of Dejagnu and
detects when a test's name includes either the source or build paths.
If any offending test names are detected then Dejagnu will print a
new result type, '# of paths in test names'.  Users should treat this
result type just like other bad results types, and aim not to increase
this number.

As well as displaying the total number of offending tests as part of
the final results, a new marker is included in both the gdb.log and
gdb.sum files, this marker starts with 'PATH: ', so an offending test
would be expected to appear like this:

  PASS: gdb.base/sometest.exp: Loaded /path/to/build/testsuite/foo.exe
  PATH: gdb.base/sometest.exp: Loaded /path/to/build/testsuite/foo.exe

This should make it easier to track down offending tests.

Currently for a local run on my machine, I don't see any offending
test names, but it is possible that different targets, or different
configurations, might currently be breaking the no paths rule.

In order to get this working I have needed to wrap two core Dejagnu
functions, log_summary, and reset_vars.  Relying on core functions
that are not part of any API is always going to be risky, given the
relatively slow rate of Dejagnu change this is probably OK for now,
and we can possibly upstream some changes to Dejagnu that would allow
this functionality to be supported in a more official way later on.

Currently if the tests are run in parallel mode the new result type is
not merged into the combined summary file so users will need to run in
non-parallel mode to check this result.  Similarly, the 'PATH: '
markers will not be merged into the combined summary file.  A later
commit will fix this.

gdb/testsuite/ChangeLog:

* lib/gdb.exp: Include check-test-names.exp library.
* lib/check-test-names.exp: New file.

4 years agoFix Ada value printing on PPC64
Tom Tromey [Mon, 11 May 2020 20:57:49 +0000 (14:57 -0600)] 
Fix Ada value printing on PPC64

The val_print removal patches introduced an Ada regression on PPC64
(probably any big-endian system).

The issue comes because value_field does not understand that Ada
wrapper fields can be bitfields that wrap a non-scalar type.  In this
case the value is already left-justified, so the justification done
there does the wrong thing.

Perhaps it would be good, eventually, to change value_field to
understand this case.  In the meantime this implements an Ada-specific
solution.

gdb/ChangeLog
2020-05-11  Tom Tromey  <tromey@adacore.com>

* ada-lang.c (ada_value_primitive_field): Now public.
* ada-lang.h (ada_value_primitive_field): Declare.
* ada-valprint.c (print_field_values): Use
ada_value_primitive_field for wrapper fields.

4 years ago[gdb/testsuite] Change kfail into xfail in gdb.ada/packed_tagged.exp
Tom de Vries [Mon, 11 May 2020 20:30:18 +0000 (22:30 +0200)] 
[gdb/testsuite] Change kfail into xfail in gdb.ada/packed_tagged.exp

Test-case gdb.ada/packed_tagged.exp contains a kfail:
...
setup_kfail "gnat compiler bug" *-*-*
...

Kfails are used to indicate problems in gdb, xfails are used to indicate
problems in the environment.

A bug in the gnat compiler is a problem in the environment rather than gdb.

Fix this by changing the kfail into an xfail.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-11  Tom de Vries  <tdevries@suse.de>

* gdb.ada/packed_tagged.exp: Change kfail into xfail.

4 years agoFix gdb.ada/attr_ref_and_charlit.exp typo
Keith Seitz [Mon, 11 May 2020 19:22:25 +0000 (12:22 -0700)] 
Fix gdb.ada/attr_ref_and_charlit.exp typo

... introduced by my last commit:

-gdb_test "print s'last"  " = 3"
+    gdb_test "print s'last"  " = 3Z

gdb/testsuite/ChangeLog
2020-05-11  Keith Seitz  <keiths@redhat.com>

* gdb.ada/attr_ref_and_charlit.exp: Fix typo.

4 years ago[gdb/testsuite] Fix gdb.cp/cpexprs-debug-types.exp inclusion
Tom de Vries [Mon, 11 May 2020 18:36:11 +0000 (20:36 +0200)] 
[gdb/testsuite] Fix gdb.cp/cpexprs-debug-types.exp inclusion

When running tests using RUNTESTFLAGS="cpexprs.exp cpexprs-debug-types.exp",
we have:
...
Running src/gdb/testsuite/gdb.cp/cpexprs-debug-types.exp ...
Running src/gdb/testsuite/gdb.cp/cpexprs.exp ...
...

In the first test-case, we have -fdebug-types-section as expected:
...
Running src/gdb/testsuite/gdb.cp/cpexprs-debug-types.exp ...
g++ -fno-stack-protector -fdiagnostics-color=never \
  -fdebug-types-section \
  -c -g \
  -o outputs/gdb.cp/cpexprs-debug-types/cpexprs-debug-types0.o \
  src/gdb/testsuite/gdb.cp/cpexprs.cc
...
but in the second test-case, we have also have -fdebug-types-section:
...
Running src/gdb/testsuite/gdb.cp/cpexprs.exp ...
g++ -fno-stack-protector -fdiagnostics-color=never \
  -fdebug-types-section \
  -c -g -g \
  -o outputs/gdb.cp/cpexprs/cpexprs0.o \
  src/gdb/testsuite/gdb.cp/cpexprs.cc
...

This is due to using a global variable flags, which is set in
cpexprs-debug-types.exp and tested for existence in cpexprs.exp.

Fix this by using a more robust inclusion mechanism, that is:
- move the bulk of the test-case cpexprs.exp to cpexprs.exp.in,
- include it from cpexprs.exp and cpexprs-debug-types.exp, and
- set flags in both .exp files

gdb/testsuite/ChangeLog:

2020-05-11  Tom de Vries  <tdevries@suse.de>

* gdb.cp/cpexprs.exp: Move everything except flags setting ...
* gdb.cp/cpexprs.exp.in: .. here.
* gdb.cp/cpexprs-debug-types.exp: Include cpexprs.exp.in instead of
cpexprs.exp.

4 years agoClean-up gdb.ada test names
Keith Seitz [Mon, 11 May 2020 17:27:32 +0000 (10:27 -0700)] 
Clean-up gdb.ada test names

This patch fixes all duplicate and tail parentheses test names.
These can really hinder automated test analysis such as used by
the buildbot.

Before:
$ cat testsuite/gdb.sum | egrep "^(PASS|FAIL|XPASS|XFAIL|KPASS|KFAIL)" \
  | sort | uniq -c | sort -n | grep -v " 1 "
      2 PASS: gdb.ada/attr_ref_and_charlit.exp: print s'last
      2 PASS: gdb.ada/bp_on_var.exp: set breakpoint pending off
      2 PASS: gdb.ada/complete.exp: complete p pck.inne
      2 PASS: gdb.ada/fun_overload_menu.exp: multiple matches for f (f (1, null))
      2 PASS: gdb.ada/type_coercion.exp: p q
      2 PASS: gdb.ada/unc_arr_ptr_in_var_rec.exp: print My_P_Object.Ptr when no longer null
      3 PASS: gdb.ada/fun_overload_menu.exp: 1

After:
<empty>

For parentheses, I've audited all occurrences of trailing parentheses.
Most offenders are of the form:

   gdb_test "p func (..)" $expected_result

I've either added a unique test name or simply removed the whitespace
between the function name and the argument list.

gdb/testsuite/ChangeLog
2020-05-11  Keith Seitz  <keiths@redhat.com>

* gdb.ada/arrayparam.exp: Resolve duplicate and tail parentheses
test names.
* gdb.ada/arrayptr.exp: Likewise.
* gdb.ada/assign_arr.exp: Likewise.
* gdb.ada/attr_ref_and_charlit.exp: Likewise.
* gdb.ada/bp_on_var.exp: Likewise.
* gdb.ada/call_pn.exp: Likewise.
* gdb.ada/complete.exp: Likewise.
* gdb.ada/fun_overload_menu.exp: Likewise.
* gdb.ada/funcall_param.exp: Likewise.
* gdb.ada/funcall_ref.exp: Likewise.
* gdb.ada/packed_array_assign.exp: Likewise.
* gdb.ada/same_component_name.exp: Likewise.
* gdb.ada/type_coercion.exp: Likewise.
* gdb.ada/unc_arr_ptr_in_var_rec.exp: Likewise.
* gdb.ada/variant_record_packed_array.exp: Likewise.

4 years ago[gdb/symtab] Save modules in .debug_names
Tom de Vries [Mon, 11 May 2020 15:13:49 +0000 (17:13 +0200)] 
[gdb/symtab] Save modules in .debug_names

When running test-case gdb.fortran/info-modules.exp with target board
debug-names, I run into:
...
FAIL: gdb.fortran/info-modules.exp: info modules: check for entry \
  'info-types-2.f90', '18', 'mod2'
...

In more detail, comparing the behaviour of the executable without and with
.debug_names section, we have:
...
-$ gdb -batch info-modules -ex "info modules"
+$ gdb -batch info-modules.debugnames -ex "info modules"
 All defined modules:

-File /data/gdb_versions/devel/src/gdb/testsuite/gdb.fortran/info-types-2.f90:
-18:     mod2
-
 File /data/gdb_versions/devel/src/gdb/testsuite/gdb.fortran/info-types.f90:
 16:     mod1
...

This is due to the fact that the .debug_names section does not contain
DW_TAG_module entries.

Fix this in debug_names::psymbol_tag.

Build and tested on x86_64-linux with target board debug-names.

gdb/ChangeLog:

2020-05-11  Tom de Vries  <tdevries@suse.de>

* dwarf2/index-write.c (debug_names::psymbol_tag): Handle
MODULE_DOMAIN.

4 years ago[gdb/symtab] Fix incomplete CU list assert in .debug_names
Tom de Vries [Mon, 11 May 2020 13:03:54 +0000 (15:03 +0200)] 
[gdb/symtab] Fix incomplete CU list assert in .debug_names

Consider the following two-file test-case:
...
$ cat main.c
extern int foo (void);

int
main (void)
{
  int sum, a, b;
  sum = a + b + foo ();
  return sum;
}
$ cat foo.c
int
foo (void)
{
  return 3;
}
...

Compiled like this:
...
$ clang-10 -gdwarf-5 -gpubnames -c main.c
$ clang-10 -gdwarf-5 -c foo.c
$ clang-10 -gdwarf-5 -gpubnames main.o foo.o
...

When loading this exec into gdb, we run into this assert:
...
$ gdb a.out
Reading symbols from a.out...

warning: Section .debug_aranges in a.out entry at offset 0 \
  debug_info_offset 0 does not exists, ignoring .debug_aranges.
src/gdb/dwarf2/read.c:6949: \
  internal-error: cutu_reader::cutu_reader(dwarf2_per_cu_data*, \
                              abbrev_table*, int, bool): \
  Assertion `this_cu->length == cu->header.get_length ()' failed.
...

The problem is that the determined length of the CU:
...
(gdb) p /x this_cu->length
$4 = 0x26a
...
does not match the actual length:
...
(gdb) p /x cu->header.get_length ()
$5 = 0x59
...

The length of the CU is determined in create_cus_from_debug_names_list, and
set based on this list in the .debug_names section:
...
  Compilation Unit offsets [
    CU[0]: 0x000000c7
  ]
...
and it is assumed that this is a complete list, so the size of the CU is
calculated using the end of the .debug_section at 0x331, making it 0x331 -
0xc7 == 0x26a.

However, the CU list is not complete:
...
$ llvm-dwarfdump -debug-info a.out \
  | grep "Compile Unit" \
  | sed 's/Compile Unit.*//'
0x00000000:
0x0000002e:
0x000000a5:
0x000000c7:
0x00000120:
0x00000157:
0x0000030f:
...
In particular, because the CU for foo.c is there at 0x120 (the rest of the CUs
is due to openSUSE having debug info for various linked in objects).

Fix the assert by not assuming to know the length of CUs in
create_cus_from_debug_names_list (if the .debug_names is not produced by GDB),
and setting it to 0, and setting it later to the actual length.

Note that this does not fix the .debug_aranges warning, that's PR25969.

Build and tested on x86_64-linux, with native and debug-names.

gdb/ChangeLog:

2020-05-11  Tom de Vries  <tdevries@suse.de>

PR symtab/25941
* dwarf2/read.c (create_cus_from_debug_names_list): Initialize CUs
with length 0, if not gdb-produced.
(cutu_reader::cutu_reader): Set CU length to actual length if 0.

gdb/testsuite/ChangeLog:

2020-05-11  Tom de Vries  <tdevries@suse.de>

PR symtab/25941
* gdb.dwarf2/clang-debug-names.exp.in: New include exp file, factored
out of ...
* gdb.dwarf2/clang-debug-names.exp: ... here.
* gdb.dwarf2/clang-debug-names-2.exp: New file.  Include
clang-debug-names.exp.in.
* gdb.dwarf2/clang-debug-names-2-foo.c: New test.
* gdb.dwarf2/clang-debug-names-2.c: New test.

4 years agoPower10 VSX scalar min-max-compare quad precision operations
Alan Modra [Mon, 11 May 2020 00:19:29 +0000 (09:49 +0930)] 
Power10 VSX scalar min-max-compare quad precision operations

opcodes/
* ppc-opc (powerpc_opcodes): Add xscmpeqqp, xscmpgeqp, xscmpgtqp,
xsmaxcqp, xsmincqp.
gas/
* testsuite/gas/ppc/scalarquad.d,
* testsuite/gas/ppc/scalarquad.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 VSX load/store rightmost element operations
Alan Modra [Mon, 11 May 2020 00:18:29 +0000 (09:48 +0930)] 
Power10 VSX load/store rightmost element operations

opcodes/
* ppc-opc.c (powerpc_opcodes): Add lxvrbx, lxvrhx, lxvrwx, lxvrdx,
stxvrbx, stxvrhx, stxvrwx, stxvrdx.
gas/
* testsuite/gas/ppc/rightmost.d,
* testsuite/gas/ppc/rightmost.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 test lsb by byte operation
Alan Modra [Mon, 11 May 2020 00:17:38 +0000 (09:47 +0930)] 
Power10 test lsb by byte operation

opcodes/
* ppc-opc.c (powerpc_opcodes): Add xvtlsbb.
gas/
* testsuite/gas/ppc/xvtlsbb.d,
* testsuite/gas/ppc/xvtlsbb.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 string operations
Alan Modra [Mon, 11 May 2020 00:16:45 +0000 (09:46 +0930)] 
Power10 string operations

opcodes/
* ppc-opc.c (powerpc_opcodes): Add vstribl, vstribr, vstrihl, vstrihr,
vclrlb, vclrrb, vstribl., vstribr., vstrihl., vstrihr..
gas/
* testsuite/gas/ppc/stringop.d,
* testsuite/gas/ppc/stringop.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 Set boolean extension
Peter Bergner [Mon, 11 May 2020 00:15:42 +0000 (09:45 +0930)] 
Power10 Set boolean extension

opcodes/
* ppc-opc.c (powerpc_opcodes) <setbc, setbcr, setnbc, setnbcr>: New
mnemonics.
gas/
* testsuite/gas/ppc/set_bool.d,
* testsuite/gas/ppc/set_bool.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 bit manipulation operations
Alan Modra [Mon, 11 May 2020 00:14:25 +0000 (09:44 +0930)] 
Power10 bit manipulation operations

opcodes/
* ppc-opc.c (UIM8, P_U8XX4_MASK): Define.
(powerpc_opcodes): Add vgnb, vcfuged, vpextd, vpdepd, vclzdm,
vctzdm, cntlzdm, pdepd, pextd, cfuged, cnttzdm.
(prefix_opcodes): Add xxeval.
gas/
* testsuite/gas/ppc/bitmanip.d,
* testsuite/gas/ppc/bitmanip.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 VSX PCV generate operations
Alan Modra [Mon, 11 May 2020 00:13:15 +0000 (09:43 +0930)] 
Power10 VSX PCV generate operations

opcodes/
* ppc-opc.c (powerpc_opcodes): Add xxgenpcvbm, xxgenpcvhm,
xxgenpcvwm, xxgenpcvdm.
gas/
* testsuite/gas/ppc/genpcv.d,
* testsuite/gas/ppc/genpcv.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 VSX Mask Manipulation Operations
Alan Modra [Mon, 11 May 2020 00:11:57 +0000 (09:41 +0930)] 
Power10 VSX Mask Manipulation Operations

opcodes/
* ppc-opc.c (MP, VXVAM_MASK): Define.
(VXVAPS_MASK): Use VXVA_MASK.
(powerpc_opcodes): Add mtvsrbmi, vexpandbm, vexpandhm, vexpandwm,
vexpanddm, vexpandqm, vextractbm, vextracthm, vextractwm,
vextractdm, vextractqm, mtvsrbm, mtvsrhm, mtvsrwm, mtvsrdm, mtvsrqm,
vcntmbb, vcntmbh, vcntmbw, vcntmbd.
gas/
* testsuite/gas/ppc/maskmanip.d,
* testsuite/gas/ppc/maskmanip.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 Reduced precision outer product operations
Alan Modra [Mon, 11 May 2020 00:10:42 +0000 (09:40 +0930)] 
Power10 Reduced precision outer product operations

include/
* opcode/ppc.h (PPC_OPERAND_ACC): Define.  Renumber following
PPC_OPERAND defines.
opcodes/
* ppc-opc.c (insert_xa6a, extract_xa6a, insert_xb6a, extract_xb6a):
New functions.
(powerpc_operands): Define ACC, PMSK8, PMSK4, PMSK2, XMSK, YMSK,
YMSK2, XA6a, XA6ap, XB6a entries.
(PMMIRR, P_X_MASK, P_XX1_MASK, P_GER_MASK): Define
(P_GER2_MASK, P_GER4_MASK, P_GER8_MASK, P_GER64_MASK): Define.
(PPCVSX4): Define.
(powerpc_opcodes): Add xxmfacc, xxmtacc, xxsetaccz,
xvi8ger4pp, xvi8ger4, xvf16ger2pp, xvf16ger2, xvf32gerpp, xvf32ger,
xvi4ger8pp, xvi4ger8, xvi16ger2spp, xvi16ger2s, xvbf16ger2pp,
xvbf16ger2, xvf64gerpp, xvf64ger, xvi16ger2, xvf16ger2np,
xvf32gernp, xvi8ger4spp, xvi16ger2pp, xvbf16ger2np, xvf64gernp,
xvf16ger2pn, xvf32gerpn, xvbf16ger2pn, xvf64gerpn, xvf16ger2nn,
xvf32gernn, xvbf16ger2nn, xvf64gernn, xvcvbf16sp, xvcvspbf16.
(prefix_opcodes): Add pmxvi8ger4pp, pmxvi8ger4, pmxvf16ger2pp,
pmxvf16ger2, pmxvf32gerpp, pmxvf32ger, pmxvi4ger8pp, pmxvi4ger8,
pmxvi16ger2spp, pmxvi16ger2s, pmxvbf16ger2pp, pmxvbf16ger2,
pmxvf64gerpp, pmxvf64ger, pmxvi16ger2, pmxvf16ger2np, pmxvf32gernp,
pmxvi8ger4spp, pmxvi16ger2pp, pmxvbf16ger2np, pmxvf64gernp,
pmxvf16ger2pn, pmxvf32gerpn, pmxvbf16ger2pn, pmxvf64gerpn,
pmxvf16ger2nn, pmxvf32gernn, pmxvbf16ger2nn, pmxvf64gernn.
gas/
* config/tc-ppc.c (pre_defined_registers): Add accumulators.
(md_assemble): Check acc specified in correct operand.
* testsuite/gas/ppc/outerprod.d,
* testsuite/gas/ppc/outerprod.s,
* testsuite/gas/ppc/vsx4.d,
* testsuite/gas/ppc/vsx4.s: New tests.
* testsuite/gas/ppc/ppc.exp: Run them.

4 years agoPower10 SIMD permute class operations
Alan Modra [Mon, 11 May 2020 00:07:14 +0000 (09:37 +0930)] 
Power10 SIMD permute class operations

opcodes/
* ppc-opc.c (insert_imm32, extract_imm32): New functions.
(insert_xts, extract_xts): New functions.
(IMM32, UIM3, IX, UIM5, SH3, XTS, P8RR): Define.
(P_XX4_MASK, P_UXX4_MASK, VSOP, P_VS_MASK, P_VSI_MASK): Define.
(VXRC_MASK, VXSH_MASK): Define.
(powerpc_opcodes): Add vinsbvlx, vsldbi, vextdubvlx, vextdubvrx,
vextduhvlx, vextduhvrx, vextduwvlx, vextduwvrx, vextddvlx,
vextddvrx, vinshvlx, vinswvlx, vinsw, vinsbvrx, vinshvrx,
vinswvrx, vinsd, vinsblx, vsrdbi, vinshlx, vinswlx, vinsdlx,
vinsbrx, vinshrx, vinswrx, vinsdrx, lxvkq.
(prefix_opcodes): Add xxsplti32dx, xxspltidp, xxspltiw, xxblendvb,
xxblendvh, xxblendvw, xxblendvd, xxpermx.
gas/
* testsuite/gas/ppc/simd_perm.d,
* testsuite/gas/ppc/simd_perm.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 128-bit binary integer operations
Alan Modra [Mon, 11 May 2020 00:06:11 +0000 (09:36 +0930)] 
Power10 128-bit binary integer operations

opcodes/
* ppc-opc.c (powerpc_opcodes): Add vrlq, vdivuq, vmsumcud, vrlqmi,
vmuloud, vcmpuq, vslq, vdivsq, vcmpsq, vrlqnm, vcmpequq, vmulosd,
vsrq, vdiveuq, vcmpgtuq, vmuleud, vsraq, vdivesq, vcmpgtsq, vmulesd,
vcmpequq., vextsd2q, vmoduq, vcmpgtuq., vmodsq, vcmpgtsq., xscvqpuqz,
xscvuqqp, xscvqpsqz, xscvsqqp, dcffixqq, dctfixqq.
gas/
* testsuite/gas/ppc/int128.d,
* testsuite/gas/ppc/int128.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 VSX 32-byte storage access
Alan Modra [Mon, 11 May 2020 00:04:49 +0000 (09:34 +0930)] 
Power10 VSX 32-byte storage access

bfd/
* elf64-ppc.c (xlate_pcrel_opt): Handle lxvp and stxvp.
opcodes/
* ppc-opc.c (insert_xtp, extract_xtp): New functions.
(XTP, DQXP, DQXP_MASK): Define.
(powerpc_opcodes): Add lxvp, stxvp, lxvpx, stxvpx.
(prefix_opcodes): Add plxvp and pstxvp.
gas/
* testsuite/gas/ppc/vsx_32byte.d,
* testsuite/gas/ppc/vsx_32byte.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.
ld/
* testsuite/ld-powerpc/pcrelopt.s: Add lxvp and stxvp.
* testsuite/ld-powerpc/pcrelopt.d: Update.

4 years agoPower10 vector integer multiply, divide, modulo insns
Alan Modra [Mon, 11 May 2020 00:02:56 +0000 (09:32 +0930)] 
Power10 vector integer multiply, divide, modulo insns

opcodes/
* ppc-opc.c (powerpc_opcodes): Add vdivuw, vdivud, vdivsw, vmulld,
vdivsd, vmulhuw, vdiveuw, vmulhud, vdiveud, vmulhsw, vdivesw,
vmulhsd, vdivesd, vmoduw, vmodud, vmodsw, vmodsd.
gas/
* testsuite/gas/ppc/vec_mul.s,
* testsuite/gas/ppc/vec_mul.d: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 byte reverse instructions
Peter Bergner [Mon, 11 May 2020 00:01:40 +0000 (09:31 +0930)] 
Power10 byte reverse instructions

opcodes/
* ppc-opc.c (powerpc_opcodes) <brd, brh, brw>: New mnemonics.
gas/
* testsuite/gas/ppc/byte_rev.d,
* testsuite/gas/ppc/byte_rev.s: New test.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPower10 Copy/Paste Extensions
Peter Bergner [Mon, 11 May 2020 00:00:19 +0000 (09:30 +0930)] 
Power10 Copy/Paste Extensions

opcodes/
* opcodes/ppc-opc.c (insert_l1opt, extract_l1opt): New functions.
(L1OPT): Define.
(powerpc_opcodes) <paste.>: Add L operand for cpu POWER10.
gas/
* testsuite/gas/ppc/power10.d: Add paste. tests.
* testsuite/gas/ppc/power10.s: Likewise.

4 years agoPower10 Add new L operand to the slbiag instruction
Peter Bergner [Sun, 10 May 2020 23:57:48 +0000 (09:27 +0930)] 
Power10 Add new L operand to the slbiag instruction

opcodes/
* ppc-opc.c (powerpc_opcodes) <slbiag>: Add variant with L operand.
gas/
* testsuite/gas/ppc/power10.s: New test.
* testsuite/gas/ppc/power10.d: Likewise.
* testsuite/gas/ppc/ppc.exp: Run it.

4 years agoPowerPC Default disassembler to -Mpower10
Alan Modra [Sun, 10 May 2020 23:55:30 +0000 (09:25 +0930)] 
PowerPC Default disassembler to -Mpower10

* ppc-dis.c (powerpc_init_dialect): Default to "power10".

4 years agoPowerPC Rename powerxx to power10
Alan Modra [Sun, 10 May 2020 23:54:14 +0000 (09:24 +0930)] 
PowerPC Rename powerxx to power10

Now that ISA3.1 is out we can finish with the powerxx silliness.

bfd/
* elf64-ppc.c: Rename powerxx to power10 throughout.
gas/
* config/tc-ppc.c (md_assemble): Update for PPC_OPCODE_POWER10
renaming.
* testsuite/gas/ppc/prefix-align.d: Use -mpower10/-Mpower10 in
place of -mfuture/-Mfuture.
* testsuite/gas/ppc/prefix-pcrel.d: Likewise.
* testsuite/gas/ppc/prefix-reloc.d: Likewise.
gold/
* powerpc.cc: Rename powerxx to power10 throughout.
include/
* elf/ppc64.h: Update comment.
* opcode/ppc.h (PPC_OPCODE_POWER10): Rename from PPC_OPCODE_POWERXX.
ld/
* testsuite/ld-powerpc/callstub-1.d: Use -mpower10/-Mpower10 in
place of -mfuture/-Mfuture.
* testsuite/ld-powerpc/notoc2.d: Likewise.
* testsuite/ld-powerpc/powerpc.exp: Likewise.
* testsuite/ld-powerpc/tlsgd.d: Likewise.
* testsuite/ld-powerpc/tlsie.d: Likewise.
* testsuite/ld-powerpc/tlsld.d: Likewise.
opcodes/
* ppc-dis.c (ppc_opts): Add "power10" entry.
(print_insn_powerpc): Update for PPC_OPCODE_POWER10 renaming.
* ppc-opc.c (POWER10): Rename from POWERXX.  Update all uses.

4 years agoUpdated French translation for the ld sub-directory and an update Spanish translation...
Nick Clifton [Mon, 11 May 2020 11:02:26 +0000 (12:02 +0100)] 
Updated French translation for the ld sub-directory and an update Spanish translation for the opcodes subdirectory.

4 years agoPR25961, buffer overflow in coff_swap_aux_in
Alan Modra [Mon, 11 May 2020 08:30:31 +0000 (18:00 +0930)] 
PR25961, buffer overflow in coff_swap_aux_in

PR 25961
* coffgen.c (coff_get_normalized_symtab): Check that buffer
contains required number of auxents before processing any auxent.
* coffswap.h (coff_swap_aux_in <C_FILE>): Only swap in extended
file name from auxents for PE.

4 years agoAutomatic date update in version.in
GDB Administrator [Mon, 11 May 2020 00:00:07 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years agogdb/fortran: Allow Flang MAIN_ in Fortran testing
Alok Kumar Sharma [Sun, 10 May 2020 19:12:00 +0000 (00:42 +0530)] 
gdb/fortran: Allow Flang MAIN_ in Fortran testing

Name of fortran main function for Flang compiler is MAIN_ while
for gfortran it is MAIN__ . In test cases MAIN__ is hardcoded for
the purpose of inserting breakpoint.

New proc is added to detect main function name depending on the
compiler used.
Fortran specific version of runto_main named fortran_runto_main
is added.

This commit adds support for Flang main function, there should be
no change for gfortran.

gdb/testsuite/ChangeLog

* lib/fortran.exp (fortran_main): New Proc, handle flang MAIN_,
(fortran_runto_main): New Proc, fortran version of runto_main.
* gdb.fortran/array-bounds-high.exp: Handle flang MAIN_.
* gdb.fortran/array-bounds.exp: Likewise.
* gdb.fortran/array-slices.exp: Likewise.
* gdb.fortran/block-data.exp: Likewise.
* gdb.fortran/charset.exp: Likewise.
* gdb.fortran/common-block.exp: Likewise.
* gdb.fortran/complex.exp: Likewise.
* gdb.fortran/derived-type-function.exp: Likewise.
* gdb.fortran/derived-type.exp: Likewise.
* gdb.fortran/info-modules.exp: Likewise.
* gdb.fortran/info-types.exp: Likewise.
* gdb.fortran/intrinsics.exp: Likewise.
* gdb.fortran/library-module.exp: Likewise.
* gdb.fortran/logical.exp: Likewise.
* gdb.fortran/max-depth.exp: Likewise.
* gdb.fortran/module.exp: Likewise.
* gdb.fortran/multi-dim.exp: Likewise.
* gdb.fortran/nested-funcs.exp: Likewise.
* gdb.fortran/print-formatted.exp: Likewise.
* gdb.fortran/print_type.exp: Likewise.
* gdb.fortran/printing-types.exp: Likewise.
* gdb.fortran/ptr-indentation.exp: Likewise.
* gdb.fortran/ptype-on-functions.exp: Likewise.
* gdb.fortran/subarray.exp: Likewise.
* gdb.fortran/vla-alloc-assoc.exp: Likewise.
* gdb.fortran/vla-datatypes.exp: Likewise.
* gdb.fortran/vla-history.exp: Likewise.
* gdb.fortran/vla-ptr-info.exp: Likewise.
* gdb.fortran/vla-ptype-sub.exp: Likewise.
* gdb.fortran/vla-ptype.exp: Likewise.
* gdb.fortran/vla-sizeof.exp: Likewise.
* gdb.fortran/vla-type.exp: Likewise.
* gdb.fortran/vla-value-sub-arbitrary.exp: Likewise.
* gdb.fortran/vla-value-sub-finish.exp: Likewise.
* gdb.fortran/vla-value-sub.exp: Likewise.
* gdb.fortran/vla-value.exp: Likewise.
* gdb.fortran/whatis_type.exp: Likewise.
* gdb.mi/mi-var-child-f.exp: Likewise.

4 years agoAutomatic date update in version.in
GDB Administrator [Sun, 10 May 2020 00:00:15 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years ago[gdb] Fix catch throw regexp matching
Tom de Vries [Sat, 9 May 2020 18:17:10 +0000 (20:17 +0200)] 
[gdb] Fix catch throw regexp matching

When running test-case gdb.mi/mi-catch-cpp-exceptions.exp, we have:
...
FAIL: gdb.mi/mi-catch-cpp-exceptions.exp: all with invalid regexp: run until \
  breakpoint in main (unknown output after running)
...

This is a regression since commit 596dc4adff "Speed up psymbol reading by
removing a copy".

Before that commit, we have:
...
$ gdb \
    -batch \
    ./outputs/gdb.mi/mi-catch-cpp-exceptions/mi-catch-cpp-exceptions \
    -ex "break 67" \
    -ex "catch throw -r blahblah" \
    -ex r
Breakpoint 1 at 0x4008e5: file mi-catch-cpp-exceptions.cc, line 67.
Catchpoint 2 (throw)

Breakpoint 1, main () at mi-catch-cpp-exceptions.cc:67
67                  return 1;   /* Stop here.  */
...
In other words:
- we set a breakpoint somewhere in main,
- we set a catchpoint with a regexp that is intended to not match any
  exception, and
- run to the breakpoint, without the catchpoint triggering.

After the commit, we have:
...
$ gdb \
    -batch \
    ./outputs/gdb.mi/mi-catch-cpp-exceptions/mi-catch-cpp-exceptions \
    -ex "break 67" \
    -ex "catch throw -r blahblah" \
    -ex r
Breakpoint 1 at 0x4008e5: file mi-catch-cpp-exceptions.cc, line 67.
Catchpoint 2 (throw)

Catchpoint 2 (exception thrown), 0x00007ffff7ab037e in __cxa_throw () from \
  /usr/lib64/libstdc++.so.6
...
In other words, the catchpoint triggers.

This is caused by this bit of the commit:
...
       type_name = cplus_typename_from_type_info (typeinfo_arg);

       canon = cp_canonicalize_string (type_name.c_str ());
-      if (!canon.empty ())
-       std::swap (type_name, canon);
+      name = (canon == nullptr
+       ? canon.get ()
+       : type_name.c_str ());
     }
   catch (const gdb_exception_error &e)
     {
       exception_print (gdb_stderr, e);
     }

-  if (!type_name.empty ())
+  if (name != nullptr)
     {
-      if (self->pattern->exec (type_name.c_str (), 0, NULL, 0) != 0)
+      if (self->pattern->exec (name, 0, NULL, 0) != 0)
...

Before the commit, we have:
- type_name == "my_exception"
- canon = ""
and the !type_name.empty () test succeeds, and gdb executes the
self->pattern->exec call.

After the commit, we have:
- type_name == "my_exception"
- canon == NULL
- name == NULL
and the name != nullptr test fails, and gdb doesn't execute the
self->pattern->exec call.

Fix this by inverting the condition for the calculation of name:
...
-      name = (canon == nullptr
+      name = (canon != nullptr
...

Build and tested on x86_64-linux.

gdb/ChangeLog:

2020-05-09  Tom de Vries  <tdevries@suse.de>

PR gdb/25955
* break-catch-throw.c (check_status_exception_catchpoint): Fix name
calculation.

4 years agoChange server_command to bool
Tom Tromey [Sat, 9 May 2020 18:04:58 +0000 (12:04 -0600)] 
Change server_command to bool

I noticed that "server_command" is an int, but really it should be a
bool.

gdb/ChangeLog
2020-05-09  Tom Tromey  <tom@tromey.com>

* top.c (server_command): Now bool.
* top.h (server_command): Now bool.

4 years agoFix for the complaint observed when symbol reading due to unsupported .debug_names...
nitachra [Sat, 9 May 2020 08:03:51 +0000 (10:03 +0200)] 
Fix for the complaint observed when symbol reading due to unsupported .debug_names form

Following complaint is observed with the executable compiled with -gdwarf-5
and -gpubnames flags - "During symbol reading: Unsupported .debug_names form
DW_FORM_ref4".  This is the form corresponding to DW_IDX_die_offset attribute.
This patch fixes this complaint.  Tested with clang 10.0.0.  Test case used -

int main()
{
int sum,a,b;
sum = a + b;
return sum;
}

clang -gdwarf-5 -gpubnames test.c -o test.out

gdb -q test.out -ex "set complaints 1" -ex "start"

Reading symbols from test.out...
During symbol reading: Unsupported .debug_names form DW_FORM_ref4 \
  [in module test.out]
Temporary breakpoint 1 at 0x400484
Starting program: test.out
During symbol reading: Unsupported .debug_names form DW_FORM_ref4 \
  [in module test.out]
During symbol reading: Unsupported .debug_names form DW_FORM_ref4 \
  [in module test.out]
During symbol reading: Unsupported .debug_names form DW_FORM_ref4 \
  [in module test.out]

gdb/dwarf2/ChangeLog:

2020-05-09  Nitika Achra  <Nitika.Achra@amd.com>

PR symtab/25952
* read.c (dw2_debug_names_iterator::next): Handle DW_FORM_ref*
and DW_IDX_die_offset.  If there is no compilation unit attribute in
the index entry, then there is a single CU.  Return the CU at O index
of compilation unit vector.

gdb/testsuite/ChangeLog:

2020-05-09  Tom de Vries  <tdevries@suse.de>

* gdb.dwarf2/clang-debug-names.exp: Remove PR25952 kfail.

4 years agoAutomatic date update in version.in
GDB Administrator [Sat, 9 May 2020 00:00:12 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years agoDon't re-process a DIE in read_lexical_block_scope
Tom Tromey [Fri, 8 May 2020 20:26:11 +0000 (14:26 -0600)] 
Don't re-process a DIE in read_lexical_block_scope

A customer reported a crash in the DWARF reader.

Investigation showed that the crash occurred in an unusual scenario: a
function was lexically scoped within some other function -- but the
inner function inlined the outer function and referred to its DIE via
DW_AT_abstract_origin.  With the executable in question,
inherit_abstract_dies could eventually call read_lexical_block_scope,
which in turn could recurse into process_die, to process a DIE that
was already being read, triggering an assert.

This came up once before; see:

https://www.sourceware.org/ml/gdb-patches/2014-02/msg00652.html

However, in this case, I don't have an easy way to reproduce.  So,
there is no test case.

I did experiment with the failing executable.  This patch fixes the
bug and doesn't seem to cause other issues.  For example, I can still
set breakpoints on the relevant functions.

gdb/ChangeLog
2020-05-08  Tom Tromey  <tromey@adacore.com>

* dwarf2/read.c (read_lexical_block_scope): Don't process a DIE
already being processed.

4 years agoMore C++-ification for struct display
Tom Tromey [Fri, 8 May 2020 20:21:22 +0000 (14:21 -0600)] 
More C++-ification for struct display

This changes displays to have a constructor, use bool and std::string,
and to be stored using std::vector.  The ALL_DISPLAYS and
ALL_DISPLAYS_SAFE macros are removed.  While internal iteration is
still done via map_display_numbers, this is updated to use a
function_view.  These changes simplify the code somewhat; for example,
free_display can now be removed in favor of ordinary destruction.

gdb/ChangeLog
2020-05-08  Tom Tromey  <tom@tromey.com>

* printcmd.c (struct display) <next>: Remove.
<display>: New constructor.
<exp_string>: Now a std::string.
<enabled_p>: Now a bool.
(display_number): Move definition earlier.
(displays): Rename from display_chain.  Now a std::vector.
(ALL_DISPLAYS, ALL_DISPLAYS_SAFE): Remove.
(display_command): Update.
(do_one_display, disable_display)
(enable_disable_display_command, do_enable_disable_display):
Update.
(free_display): Remove.
(clear_displays): Rewrite.
(delete_display): Update.
(map_display_numbers): Use function_view.  Remove "data"
parameter.  Update.
(do_delete_display): Remove.
(undisplay_command): Update.
(do_one_display, do_displays, disable_display)
(info_display_command): Update.
(do_enable_disable_display): Remove.
(enable_disable_display_command)
(clear_dangling_display_expressions): Update.

4 years agoRemove ALL_PSPACES
Tom Tromey [Fri, 8 May 2020 20:21:22 +0000 (14:21 -0600)] 
Remove ALL_PSPACES

This removes the ALL_PSPACES macro.  In this case it seemed cleanest
to change how program spaces are stored -- instead of using a linked
list, they are now stored in a std::vector.

gdb/ChangeLog
2020-05-08  Tom Tromey  <tom@tromey.com>

* symtab.c (set_symbol_cache_size)
(maintenance_print_symbol_cache, maintenance_flush_symbol_cache)
(maintenance_print_symbol_cache_statistics): Update.
* symmisc.c (print_symbol_bcache_statistics)
(print_objfile_statistics, maintenance_print_objfiles)
(maintenance_info_symtabs, maintenance_check_symtabs)
(maintenance_expand_symtabs, maintenance_info_line_tables):
Update.
* symfile-debug.c (set_debug_symfile): Update.
* source.c (forget_cached_source_info): Update.
* python/python.c (gdbpy_progspaces): Update.
* psymtab.c (maintenance_info_psymtabs): Update.
* probe.c (parse_probes): Update.
* linespec.c (iterate_over_all_matching_symtabs)
(collect_symtabs_from_filename, search_minsyms_for_name): Update.
* guile/scm-progspace.c (gdbscm_progspaces): Update.
* exec.c (exec_target::close): Update.
* ada-tasks.c (ada_tasks_new_objfile_observer): Update.
* breakpoint.c (print_one_breakpoint_location)
(create_longjmp_master_breakpoint)
(create_std_terminate_master_breakpoint): Update.
* progspace.c (program_spaces): Now a std::vector.
(maybe_new_address_space): Update.
(add_program_space): Remove.
(program_space::program_space): Update.
(remove_program_space): Update.
(number_of_program_spaces): Remove.
(print_program_space, update_address_spaces): Update.
* progspace.h (program_spaces): Change type.
(ALL_PSPACES): Remove.
(number_of_program_spaces): Don't declare.
(struct program_space) <next>: Remove.

4 years agoRemove ALL_SO_LIBS and so_list_head
Tom Tromey [Fri, 8 May 2020 20:21:22 +0000 (14:21 -0600)] 
Remove ALL_SO_LIBS and so_list_head

This patch started as an attempt to replace ALL_SO_LIBS with an
ordinary C++ iterator.  However, then I tripped over the so_list_head
define again, and decided to remove it as well.

gdb/ChangeLog
2020-05-08  Tom Tromey  <tom@tromey.com>

* mi/mi-cmd-file.c (mi_cmd_file_list_shared_libraries): Update.
* solib-svr4.c (svr4_fetch_objfile_link_map): Update.
(enable_break): Update.
* solib-frv.c (frv_fdpic_find_global_pointer): Update.
(frv_fdpic_find_canonical_descriptor): Update.
(frv_fetch_objfile_link_map): Update.
* progspace.c (program_space::free_all_objfiles): Update.
(program_space::solibs): New method.
* progspace.h (struct program_space) <solibs>: New method.
* solist.h (master_so_list): Don't declare.
(ALL_SO_LIBS): Remove.
* solib.h (so_list_head): Remove.
(update_solib_list): Update comment.
* solib.c (master_so_list): Remove.
(solib_used, update_solib_list, solib_add)
(info_sharedlibrary_command, clear_solib)
(reload_shared_libraries_1, remove_user_added_objfile): Update.

4 years agoRemove ALL_EXTENSION_LANGUAGES and ALL_ENABLED_EXTENSION_LANGUAGES
Tom Tromey [Fri, 8 May 2020 20:21:22 +0000 (14:21 -0600)] 
Remove ALL_EXTENSION_LANGUAGES and ALL_ENABLED_EXTENSION_LANGUAGES

This removes the ALL_EXTENSION_LANGUAGES and
ALL_ENABLED_EXTENSION_LANGUAGES macros, in favor of ordinary
iterators.  For ALL_ENABLED_EXTENSION_LANGUAGES, I chose to simply
inline the check, as that seemed simpler than trying to make
filtered_iterator work for std::array.  (As an aside, this sort of
thing will be easier once we can use the ranges library...)

gdb/ChangeLog
2020-05-08  Tom Tromey  <tom@tromey.com>

* extension.c (extension_languages): Now a std::array.
(ALL_EXTENSION_LANGUAGES): Remove.
(get_ext_lang_defn, get_ext_lang_of_file)
(eval_ext_lang_from_control_command): Update.
(finish_ext_lang_initialization)
(auto_load_ext_lang_scripts_for_objfile)
(ext_lang_type_printers::ext_lang_type_printers)
(apply_ext_lang_type_printers)
(ext_lang_type_printers::~ext_lang_type_printers)
(apply_ext_lang_val_pretty_printer, apply_ext_lang_frame_filter)
(preserve_ext_lang_values, get_breakpoint_cond_ext_lang)
(breakpoint_ext_lang_cond_says_stop, check_quit_flag)
(get_matching_xmethod_workers, ext_lang_colorize)
(ext_lang_before_prompt): Update.
(ALL_ENABLED_EXTENSION_LANGUAGES): Remove.

4 years agoSpeed up psymbol reading by removing a copy
Tom Tromey [Fri, 8 May 2020 20:14:05 +0000 (14:14 -0600)] 
Speed up psymbol reading by removing a copy

I noticed that cp_canonicalize_string and friends copy a
unique_xmalloc_ptr to a std::string.  However, this copy isn't
genuinely needed anywhere, and it serves to slow down DWARF psymbol
reading.

This patch removes the copy and updates the callers to adapt.

This speeds up the reader from 1.906 seconds (mean of 10 runs, of gdb
on a copy of itself) to 1.888 seconds (mean of 10 runs, on the same
copy as the first trial).

gdb/ChangeLog
2020-05-08  Tom Tromey  <tom@tromey.com>

* symtab.h (class demangle_result_storage) <set_malloc_ptr>: New
overload.
<swap_string, m_string>: Remove.
* symtab.c (demangle_for_lookup, completion_list_add_symbol):
Update.
* stabsread.c (define_symbol, read_type): Update.
* linespec.c (find_linespec_symbols): Update.
* gnu-v3-abi.c (gnuv3_get_typeid): Update.
* dwarf2/read.c (dwarf2_canonicalize_name): Update.
* dbxread.c (read_dbx_symtab): Update.
* cp-support.h (cp_canonicalize_string_full)
(cp_canonicalize_string, cp_canonicalize_string_no_typedefs):
Return unique_xmalloc_ptr.
* cp-support.c (inspect_type): Update.
(cp_canonicalize_string_full): Return unique_xmalloc_ptr.
(cp_canonicalize_string_no_typedefs, cp_canonicalize_string):
Likewise.
* c-typeprint.c (print_name_maybe_canonical): Update.
* break-catch-throw.c (check_status_exception_catchpoint):
Update.

4 years ago[gdb] Fix stepping over fork with follow-fork-mode child and gcc-8
Tom de Vries [Fri, 8 May 2020 15:26:32 +0000 (17:26 +0200)] 
[gdb] Fix stepping over fork with follow-fork-mode child and gcc-8

When running test-case gdb.threads/fork-child-threads.exp with gcc-8 instead
of gcc-7, we have:
...
 (gdb) next^M
 [Attaching after Thread 0x7ffff7fae740 (LWP 27574) fork to child process \
   27578]^M
 [New inferior 2 (process 27578)]^M
 [Detaching after fork from parent process 27574]^M
 [Inferior 1 (process 27574) detached]^M
 [Thread debugging using libthread_db enabled]^M
 Using host libthread_db library "/lib64/libthread_db.so.1".^M
 [Switching to Thread 0x7ffff7fae740 (LWP 27578)]^M
-main () at src/gdb/testsuite/gdb.threads/fork-child-threads.c:41^M
+main () at src/gdb/testsuite/gdb.threads/fork-child-threads.c:34^M
-41            i = pthread_create (&thread, NULL, start, NULL);^M
+34        switch (fork ())^M
-(gdb) PASS: gdb.threads/fork-child-threads.exp: next over fork
+(gdb) FAIL: gdb.threads/fork-child-threads.exp: next over fork
...

This is due to the fact that gcc-8 generates more precise line info, making
the instruction after the call to fork a "recommended breakpoint location".
However, it is a bug because next is supposed to move to the next source
line.

The problem is that in process_event_stop_test we hit this code:
...
  if ((ecs->event_thread->suspend.stop_pc == stop_pc_sal.pc)
      && (ecs->event_thread->current_line != stop_pc_sal.line
  || ecs->event_thread->current_symtab != stop_pc_sal.symtab))
    {
      if (stop_pc_sal.is_stmt)
{
  /* We are at the start of a different line.  So stop.  Note that
     we don't stop if we step into the middle of a different line.
     That is said to make things like for (;;) statements work
     better.  */
  if (debug_infrun)
    fprintf_unfiltered (gdb_stdlog,
"infrun: stepped to a different line\n");
  end_stepping_range (ecs);
  return;
}
...
because current_line and current_symtab have initial values:
...
(gdb) p ecs->event_thread->current_line
$8 = 0
(gdb) p ecs->event_thread->current_symtab
$9 = (symtab *) 0x0
...

Fix this in follow_fork by copying current_line and current_symtab from
parent thread to child thread.

Tested on x86_64-linux, with gcc 7.5.0 and gcc 10.0.1.

gdb/ChangeLog:

2020-05-08  Tom de Vries  <tdevries@suse.de>

* infrun.c (follow_fork): Copy current_line and current_symtab to
child thread.

4 years ago[gdb/testsuite] Add gdb.dwarf2/clang-debug-names.c
Tom de Vries [Fri, 8 May 2020 14:24:09 +0000 (16:24 +0200)] 
[gdb/testsuite] Add gdb.dwarf2/clang-debug-names.c

Add test-case with .debug_names section using DW_FORM_ref4.

There's currently no support for .debug_names in the dwarf assembler, so we
use plain _emit rather than something more structured.

Consequently, we cannot use regular declare_labels-generated labels to refer
from .debug_names to .debug_info.  Instead, we use labels with a
predefined name, which we generate using _compute_label, and then define using
define_label.

This is the test-case for PR25952, so kfail the corresponding test.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-08  Tom de Vries  <tdevries@suse.de>

* gdb.dwarf2/clang-debug-names.c: New test.
* gdb.dwarf2/clang-debug-names.exp: New file.

4 years agoAutomatic date update in version.in
GDB Administrator [Fri, 8 May 2020 00:00:14 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years agogdb: small cleanup of async-event.c structs
Simon Marchi [Thu, 7 May 2020 15:41:45 +0000 (11:41 -0400)] 
gdb: small cleanup of async-event.c structs

This is a small cleanup to normalize the structures in async-event.c
with the rest of the code base:

- Remove the unnecessary typedefs
- Fix indentation of struct bodies
- Put comments above fields

No functional changes expected.

gdb/ChangeLog:

* async-event.c (struct async_signal_handler, struct
async_event_handler): Reformat, remove typedef.

4 years agogdb: remove TYPE_DYN_PROP_LIST macro
Simon Marchi [Thu, 7 May 2020 15:18:42 +0000 (11:18 -0400)] 
gdb: remove TYPE_DYN_PROP_LIST macro

Remove this macro, which abstracts how to obtain the dyn_prop_list of a
given type.  We could replace it with a method on `struct type`, but I
don't think it's needed, as the only code that accesses the dynamic prop
list directly is internal gdbtypes.c code (that can be seen as code
internal to `struct type`).  So it can just refer to the field directly.

gdb/ChangeLog:

* gdbtypes.h (TYPE_DYN_PROP_LIST): Remove.  Update all users
access thistype->main_type->dyn_prop_list directly.

4 years agogdb: make remove_dyn_prop a method of struct type
Simon Marchi [Thu, 7 May 2020 15:08:54 +0000 (11:08 -0400)] 
gdb: make remove_dyn_prop a method of struct type

Move remove_dyn_prop, currently a free function, to be a method of
struct type.

gdb/ChangeLog:

* gdbtypes.h (struct type) <remove_dyn_prop>: New method.
(remove_dyn_prop): Remove.  Update all users to use
type::remove_dyn_prop.
* gdbtypes.c (remove_dyn_prop): Rename to...
(type::remove_dyn_prop): ... this.

4 years agogdb: make add_dyn_prop a method of struct type
Simon Marchi [Thu, 7 May 2020 15:17:33 +0000 (11:17 -0400)] 
gdb: make add_dyn_prop a method of struct type

Move add_dyn_prop, currently a free function, to be a method of struct
type.

gdb/ChangeLog:

* gdbtypes.h (struct type) <add_dyn_prop>: New method.
(add_dyn_prop): Remove.  Update all users to use
type::add_dyn_prop.
* gdbtypes.c (add_dyn_prop): Rename to...
(type::add_dyn_prop): ... this.

4 years agogdb: make get_dyn_prop a method of struct type
Simon Marchi [Thu, 7 May 2020 15:32:25 +0000 (11:32 -0400)] 
gdb: make get_dyn_prop a method of struct type

Move get_dyn_prop, currently a free function, to be a method on struct
type.

gdb/ChangeLog:

* gdbtypes.h (struct type) <get_dyn_prop>: New method.
(get_dyn_prop): Remove.  Update all users to use
type::dyn_prop.
* gdbtypes.c (get_dyn_prop): Rename to...
(type::dyn_prop): ... this.

4 years agoobjcopy: Mention 'entry address' in description
Andrew Burgess [Tue, 5 May 2020 14:48:13 +0000 (15:48 +0100)] 
objcopy: Mention 'entry address' in description

The ELF header contains a start address, which is also sometimes
called the entry address.  The 'objdump -x' output calls this field
the 'start address', while readelf calls it 'entry point address'.
The linker talks about setting the 'entry' point in its manual.

I've always thought of this field as the 'entry address', and so when
I recently wanted to know if objcopy could adjust this field I opened
up the manual and searched for 'entry', which found no hits.

I thought it would be useful (for people like me) if the description
of 'set-start' and 'adjust-start' in the objcopy manual mentioned the
word 'entry' to make it easier to find.

binutils/ChangeLog:

* doc/binutils.texi: Mention 'entry address' in the set-start and
adjust-start options descriptions.

4 years agoUpdated French translation for the gprof sub-directory.
Nick Clifton [Thu, 7 May 2020 09:48:21 +0000 (10:48 +0100)] 
Updated French translation for the gprof sub-directory.

4 years agoAutomatic date update in version.in
GDB Administrator [Thu, 7 May 2020 00:00:12 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years agogdb: remove main_type::flag_static
Simon Marchi [Wed, 6 May 2020 16:26:05 +0000 (12:26 -0400)] 
gdb: remove main_type::flag_static

It is not used.

gdb/ChangeLog:

* gdbtypes.h (struct main_type) <flag_static>: Remove.

4 years agogdb: handle endbr64 instruction in amd64_analyze_prologue
Simon Marchi [Wed, 6 May 2020 16:01:37 +0000 (12:01 -0400)] 
gdb: handle endbr64 instruction in amd64_analyze_prologue

v2:
  - test: build full executable instead of object
  - test: add and use supports_fcf_protection
  - test: use gdb_test_multiple's -wrap option
  - test: don't execute gdb_assert if failed to get breakpoint address

Some GCCs now enable -fcf-protection by default.  This is the case, for
example, with GCC 9.3.0 on Ubuntu 20.04.  Enabling it causes the
`endbr64` instruction to be inserted at the beginning of all functions
and that breaks GDB's prologue analysis.

I noticed this because it gives many failures in gdb.base/break.exp.
But let's take this dummy program and put a breakpoint on main:

    int main(void)
    {
        return 0;
    }

Without -fcf-protection, the breakpoint is correctly put after the prologue:

    $ gcc test.c -g3 -O0 -fcf-protection=none
    $ ./gdb -q -nx --data-directory=data-directory a.out
    Reading symbols from a.out...
    (gdb) disassemble main
    Dump of assembler code for function main:
       0x0000000000001129 <+0>:     push   %rbp
       0x000000000000112a <+1>:     mov    %rsp,%rbp
       0x000000000000112d <+4>:     mov    $0x0,%eax
       0x0000000000001132 <+9>:     pop    %rbp
       0x0000000000001133 <+10>:    retq
    End of assembler dump.
    (gdb) b main
    Breakpoint 1 at 0x112d: file test.c, line 3.

With -fcf-protection, the breakpoint is incorrectly put on the first
byte of the function:

    $ gcc test.c -g3 -O0 -fcf-protection=full
    $ ./gdb -q -nx --data-directory=data-directory a.out
    Reading symbols from a.out...
    (gdb) disassemble main
    Dump of assembler code for function main:
       0x0000000000001129 <+0>:     endbr64
       0x000000000000112d <+4>:     push   %rbp
       0x000000000000112e <+5>:     mov    %rsp,%rbp
       0x0000000000001131 <+8>:     mov    $0x0,%eax
       0x0000000000001136 <+13>:    pop    %rbp
       0x0000000000001137 <+14>:    retq
    End of assembler dump.
    (gdb) b main
    Breakpoint 1 at 0x1129: file test.c, line 2.

Stepping in amd64_skip_prologue, we can see that the prologue analysis,
for GCC-compiled programs, is done in amd64_analyze_prologue by decoding
the instructions and looking for typical patterns.  This patch changes
the analysis to check for a prologue starting with the `endbr64`
instruction, and skip it if it's there.

gdb/ChangeLog:

* amd64-tdep.c (amd64_analyze_prologue): Check for `endbr64`
instruction, skip it if it's there.

gdb/testsuite/ChangeLog:

* gdb.arch/amd64-prologue-skip-cf-protection.exp: New file.
* gdb.arch/amd64-prologue-skip-cf-protection.c: New file.

4 years agoUpdated Swedish translation for the gas sub-directory
Nick Clifton [Wed, 6 May 2020 13:17:36 +0000 (14:17 +0100)] 
Updated Swedish translation for the gas sub-directory

4 years ago[gdb/testsuite] Fix gdb.reverse/consecutive-{precsave,reverse}.exp with gcc-8
Tom de Vries [Wed, 6 May 2020 12:48:50 +0000 (14:48 +0200)] 
[gdb/testsuite] Fix gdb.reverse/consecutive-{precsave,reverse}.exp with gcc-8

When running test-cases gdb.reverse/consecutive-precsave.exp and
gdb.reverse/consecutive-reverse.exp with gcc-8, we get:
...
FAIL: gdb.reverse/consecutive-precsave.exp: stopped at bp, 2nd instr
FAIL: gdb.reverse/consecutive-reverse.exp: stopped at bp, 2nd instr
...

These FAILs are duplicates of the FAILs fixed in commit 7c99e7e2b08
"[gdb/testsuite] Fix gdb.base/consecutive.exp with gcc-8".

Fix these in the same manner.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-06  Tom de Vries  <tdevries@suse.de>

* gdb.reverse/consecutive-precsave.exp: Handle if instruction after
breakpoint is at a "recommended breakpoint location".
* gdb.reverse/consecutive-reverse.exp: Same.

4 years ago[gdb/testsuite] Fix gdb.base/watchpoint-reuse-slot.exp with gcc-8
Tom de Vries [Wed, 6 May 2020 12:27:36 +0000 (14:27 +0200)] 
[gdb/testsuite] Fix gdb.base/watchpoint-reuse-slot.exp with gcc-8

When running test-case gdb.base/watchpoint-reuse-slot.exp with gcc-8 instead
of gcc-7, we have:
...
 (gdb) PASS: $conf: watch *(buf.byte + 0 + 0)@1
 stepi^M
-0x00000000004004b9      34        for (i = 0; i < 100000; i++);^M
+34        for (i = 0; i < 100000; i++);^M
-(gdb) PASS: $conf: stepi advanced
+(gdb) FAIL: $conf: stepi advanced
...
where $conf is "gdb.base/watchpoint-reuse-slot.exp: hw-watch: always-inserted
off: watch x watch: : width 1, iter 0: base + 0".

This is due to the fact that gcc-8 generates more precise line info, making
the instruction at 0x4004b9 a "recommended breakpoint location", such that gdb
no longer prints the instruction address.

Fix this by getting the instruction address by printing $pc.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-06  Tom de Vries  <tdevries@suse.de>

* gdb.base/watchpoint-reuse-slot.exp (stepi): Print $pc to get current
address.

4 years agoSection "3.1 Preprocessing" of the online GAS manual has a wrong reference to "Using...
Nick Clifton [Wed, 6 May 2020 12:21:02 +0000 (13:21 +0100)] 
Section "3.1 Preprocessing" of the online GAS manual has a wrong reference to "Using GNU CC".  This fixes that link.

PR 25927
* doc/as.texi (Preprocessing): Replace cross reference to not
existant document with a URL to the equivalent page in the GCC
manual.

4 years ago[gdb/testsuite] Fix cur_addr update in gdb.base/watchpoint-reuse-slot.exp
Tom de Vries [Wed, 6 May 2020 12:13:02 +0000 (14:13 +0200)] 
[gdb/testsuite] Fix cur_addr update in gdb.base/watchpoint-reuse-slot.exp

I noticed this code in gdb.base/watchpoint-reuse-slot.exp, proc stepi:
...
    gdb_test_multiple "stepi" $test {
-re "($hex).*[string_to_regexp $srcline]\r\n$gdb_prompt $" {
    set addr $expect_out(1,string)
    if {$addr != $cur_addr} {
pass $test
    } else {
fail $test
    }
    set cur_addr addr
}
    }
...

The variable cur_addr is documented as:
...
 # The address the program is stopped at currently.
 set cur_addr ""
...
but in the gdb_test_multiple clause we assign the string "addr" to cur_addr,
while $addr contains the current address.

Fix this by assigning $addr instead "addr".

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-06  Tom de Vries  <tdevries@suse.de>

* gdb.base/watchpoint-reuse-slot.exp: Fix incorrect assignment.

4 years ago[gdb/testsuite] Fix gdb.base/store.exp with gcc-10
Tom de Vries [Wed, 6 May 2020 11:49:34 +0000 (13:49 +0200)] 
[gdb/testsuite] Fix gdb.base/store.exp with gcc-10

When running gdb.base/store.exp with gcc-10 instead of gcc-9, we have:
...
 (gdb) PASS: gdb.base/store.exp: continue to wack_double
 print l^M
-$22 = <optimized out>^M
+$22 = -1^M
-(gdb) UNSUPPORTED: gdb.base/store.exp: var double l; print old l, expecting -1
-(gdb) PASS: gdb.base/store.exp: var double l; print old l, expecting -1
+print r^M
+$23 = <optimized out>^M
+(gdb) FAIL: gdb.base/store.exp: var double l; print old r, expecting -2
...

With gcc-9, there's no location info for both l and r, but with gcc-10,
there's location info for l, but not r.

The test-case only checks for location info availability of l, and then
assumes location info for r is also available.

Fix this by allowing missing location info for r.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-06  Tom de Vries  <tdevries@suse.de>

* gdb.base/store.exp (check_set, up_set): Allowing missing location
info for r.

4 years ago[gdb/testsuite] Fix gdb.base/shlib-call.exp with gcc-8
Tom de Vries [Wed, 6 May 2020 09:50:52 +0000 (11:50 +0200)] 
[gdb/testsuite] Fix gdb.base/shlib-call.exp with gcc-8

When running test-case gdb.base/shlib-call.exp with gcc-8 instead of gcc-7, we
have:
...
 (gdb) step^M
-main () at /data/gdb_versions/devel/src/gdb/testsuite/gdb.base/shmain.c:42^M
-42        g = mainshr1(g);^M
-(gdb) PASS: gdb.base/shlib-call.exp: step out of shr2 epilogue to main
+main () at /data/gdb_versions/devel/src/gdb/testsuite/gdb.base/shmain.c:41^M
+41        g = shr2(g);^M
+(gdb) FAIL: gdb.base/shlib-call.exp: step out of shr2 epilogue to main
...

This is due to the fact that gcc-8 generates more precise line info, making
the instruction after the call to shr2 at 0x4008f1:
...
  4008e4:  8b 05 aa 07 20 00  mov    0x2007aa(%rip),%eax  # 601094 <g>
  4008ea:  89 c7              mov    %eax,%edi
  4008ec:  e8 1f fe ff ff     callq  400710 <shr2@plt>
  4008f1:  89 05 9d 07 20 00  mov    %eax,0x20079d(%rip)  # 601094 <g>
...
a "recommended breakpoint location":
...
  [0x00000287]  Special opcode 187: advance Address by 13 to 0x4008f1 and \
    Line by 0 to 41
...
so when stepping out of shr2, gdb steps back onto line 41, the line containing
the call to shr2.

Fix this by detecting this situation and adding an extra step to reach
line 42.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-06  Tom de Vries  <tdevries@suse.de>

* gdb.base/shlib-call.exp: Add extra step to reach shmain.c:42, if
necessary.

4 years ago[gdb/testsuite] Fix gdb_unbuffer_output return-type
Tom de Vries [Wed, 6 May 2020 07:41:26 +0000 (09:41 +0200)] 
[gdb/testsuite] Fix gdb_unbuffer_output return-type

When running test-case gdb.base/shlib-call.exp with clang, we get:
...
gdb compile failed, In file included from shmain.c:6:
unbuffer_output.c:39:1: warning:
      control reaches end of non-void function [-Wreturn-type]
}
^
1 warning generated.
...

Fix this by changing the return-type to void.

Tested on x86_64-linux.

gdb/testsuite/ChangeLog:

2020-05-06  Tom de Vries  <tdevries@suse.de>

* lib/unbuffer_output.c (gdb_unbuffer_output): Change return type to
void.

4 years ago[gdb/testsuite] Fix gdb.base/consecutive.exp with gcc-8
Tom de Vries [Wed, 6 May 2020 05:07:47 +0000 (07:07 +0200)] 
[gdb/testsuite] Fix gdb.base/consecutive.exp with gcc-8

When running test-case gdb.base/consecutive.exp with gcc-8 instead of gcc-7,
we get:
...
 (gdb) step^M
 ^M
-Breakpoint 3, 0x00000000004004b1 in foo () at consecutive.c:10^M
+Breakpoint 3, foo () at consecutive.c:10^M
 10        return a[0] + a[1] + a[2] + a[3] + a[4] + a[5] + a[6];^M
-(gdb) PASS: gdb.base/consecutive.exp: stopped at bp, 2nd instr
+(gdb) FAIL: gdb.base/consecutive.exp: stopped at bp, 2nd instr
...

This is due to the fact that gcc-8 generates more precise line info, making
the breakpoint address a "recommended breakpoint location", and consequently
gdb doesn't print the address prefix anymore.

Fix the FAIL by checking in the test-case whether the breakpoint address is at
"recommended breakpoint location" or not.

gdb/testsuite/ChangeLog:

2020-05-06  Tom de Vries  <tdevries@suse.de>

* lib/gdb.exp (is_stmt_addresses, hex_in_list): New proc, factored out
of ...
* gdb.base/async.exp: ... here.
* gdb.base/consecutive.exp: Handle if 2nd breakpoint is at a
"recommended breakpoint location".

4 years ago[gdb/testsuite] Compile compile-ifunc.c with -Wno-attribute-alias
Tom de Vries [Wed, 6 May 2020 03:27:48 +0000 (05:27 +0200)] 
[gdb/testsuite] Compile compile-ifunc.c with -Wno-attribute-alias

Consider the following test-case:
...
$ cat 1.c
typedef int (*final_t) (int arg);
int final (int arg)
  { return arg + 1; }
final_t gnu_ifunc (void)
  { return final; }
int gnu_ifunc_alias (int) __attribute__ ((ifunc ("gnu_ifunc")));
int main (void)
  { return gnu_ifunc_alias (10); }
...
with result:
...
$ gcc 1.c
$ ./a.out; echo $?
11
...

The test-case uses the ifunc attribute, but there's another solution using
%gnu_indirect_function.  Consider 2.c and 3.c:
...
$ cat 2.c
typedef int (*final_t) (int arg);
int final (int arg)
  { return arg + 1; }
asm (".type gnu_ifunc, %gnu_indirect_function");
final_t gnu_ifunc (void)
  { return final; }
$ cat 3.c
extern int gnu_ifunc (int);
int main (void)
  { return gnu_ifunc (10); }
...

However, it can be inconvenient to have seperate files for the incompatible
decls of gnu_ifunc, so we can use this in a single file like this:
...
$ cat 4.c
typedef int (*final_t) (int arg);
int final (int arg)
  { return arg + 1; }
asm (".type gnu_ifunc, %gnu_indirect_function");
final_t gnu_ifunc (void)
  { return final; }
extern int gnu_ifunc_alias (int arg) __attribute__ ((alias ("gnu_ifunc")));
int main (void)
  { return gnu_ifunc_alias (10); }
...

This alias trick works ok at -O0, but not at -O2:
...
$ gcc 4.c
$ ./a.out; echo $?
11
$ gcc 4.c
$ ./a.out; echo $?
176
...
and produces a warning with gcc-8 and later:
...
$ gcc-8 4.c
4.c:7:12: warning: 'gnu_ifunc_alias' alias between functions of incompatible \
  types 'int(int)' and 'int (*(void))(int)' [-Wattribute-alias]
 extern int gnu_ifunc_alias (int arg) __attribute__ ((alias ("gnu_ifunc")));
            ^~~~~~~~~~~~~~~
4.c:5:9: note: aliased declaration here
 final_t gnu_ifunc (void)
         ^~~~~~~~~
...
The warning is correct, but the mismatch is intentional.

The last variant (%gnu_indirect_function + alias) is used in
gdb.compile/compile-ifunc.c, so we run into the warning with recent gcc.

Fix the warning by compiling with -Wno-attribute-alias.

Tested the test-case on x86_64-linux with gcc-10, and observed I no longer see
the warning:
...
Running src/gdb/testsuite/gdb.compile/compile-ifunc.exp ...

                === gdb Summary ===

nr of untested testcases         1
...

gdb/testsuite/ChangeLog:

2020-05-06  Tom de Vries  <tdevries@suse.de>

* gdb.compile/compile-ifunc.exp: Use -Wno-attribute-alias.

4 years agoAutomatic date update in version.in
GDB Administrator [Wed, 6 May 2020 00:00:06 +0000 (00:00 +0000)] 
Automatic date update in version.in

4 years agogdb: remove main_type::flag_incomplete
Simon Marchi [Tue, 5 May 2020 20:59:32 +0000 (16:59 -0400)] 
gdb: remove main_type::flag_incomplete

It is unused.  The corresponding macro was removed in c3236f84c17 ("gdb:
remove TYPE_INCOMPLETE").

gdb/ChangeLog:

* gdbtypes.h (struct main_type) <flag_incomplete>: Remove.

4 years agoMention the NetBSD support in "info proc" documentation
Kamil Rytarowski [Tue, 5 May 2020 13:48:17 +0000 (15:48 +0200)] 
Mention the NetBSD support in "info proc" documentation

gdb/doc/ChangeLog:

        * gdb.texinfo (info proc, info proc cmdline, info proc cwd)
        (info proc exe, info proc mappings, info proc stat): Mention
        NetBSD support.

4 years agoRestore readelf's warnings that describe real problems with the file being examined...
Nick Clifton [Tue, 5 May 2020 15:16:03 +0000 (16:16 +0100)] 
Restore readelf's warnings that describe real problems with the file being examined.  Fix bug displaying empty file name tables.

binutils* dwarf.c (do_checks): New global variable.
(display_formatted_table): Warn about an unexpected number of
columns in the table, if checks are enabled.  Do not complain
about the lack of data following the number of entries in the
table if the table is empty.
(display_debug_lines_decoded): Only warn about an unexpected
number of columns in a table if checks are enabled.
* dwarf.h (do_checks): Add a prototype.
* elfcomm.c (error): Remove weak attribute.
(warn): Likewise.
* readelf.c (do_checks): Delete.
(warn): Delete.
(process_section_headers): Only warn about empty sections if
checks are enabled.

gas * dwarf2dbg.c (out_dir_and_file_list): Add comments describing the
construction of a DWARF-5 directory name table.
* testsuite/gas/elf/pr25917.d: Update expected output.

4 years ago[GAS] change of ELF flags initial value in rx-linux
Gunther Nikl [Tue, 5 May 2020 09:19:41 +0000 (10:19 +0100)] 
[GAS] change of ELF flags initial value in rx-linux

* config/tc-rx.c (elf_flags): Initialize for non-linux targets.
(md_parse_option): Remove initialization of elf_flags.

This page took 0.053487 seconds and 4 git commands to generate.