Matthew Khouzam [Mon, 21 Sep 2015 20:06:28 +0000 (16:06 -0400)] 
segments: make TreeMapStore thread safe
This change will allow querying/iterating over the TreeMapStore is being built.
The data is not always accurate, but it is true to the time of the query.
Change-Id: I8721002fda68f7e88cb9ec742d76568a8c53e9ae
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56381
Reviewed-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Tested-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-by: Hudson CI
Alexandre Montplaisir [Fri, 10 Jul 2015 23:55:56 +0000 (19:55 -0400)] 
Add new custom test suite to run the SWTBot tests repeatedly
Change-Id: Ida815384e4aed172194b830d196f3a595a1dc8d4
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51779
Reviewed-by: Hudson CI
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Tested-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Alexandre Montplaisir [Fri, 21 Aug 2015 01:17:01 +0000 (21:17 -0400)]
lttng: Add a ILttngUstEventLayout
The event names exposed by LTTng-UST for the memory analysis have
changed in LTTng 2.7. They are now prefixed by "lttng_ust_*",
instead of the previous "ust_*".
To support various versions in parallel, we can use a "event
layout" interface, similarly to what was done for kernel traces
with IKernelEventLayout.
This patch adds a new ILttngUstEventLayout interface, and
implementations for 2.6- and 2.7+ versions. The memory
analysis itself will be moved to use this layout in the
following patch.
Change-Id: Id1f4bde4f5fc56bb3f6d003e51716944ce2c3f93
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/50684
Reviewed-by: Hudson CI
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Tested-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Patrick Tasse [Wed, 23 Sep 2015 18:01:39 +0000 (14:01 -0400)] 
tmf: Bug 478216: Incorrect text highlighting in event table due to tabs
Use GC.textExtent() instead of GC.stringExtent() to calculate the
matching text background highlight x-coordinates, as this method
supports tab expansion.
Change-Id: I98cfdc7e46e7f10086882650f941fc099544f898
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/56554
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Marc-Andre Laperle [Tue, 22 Sep 2015 18:38:37 +0000 (14:38 -0400)] 
rcp: Fix error "Could not resolve module: org.eclipse.equinox.ds"
org.eclipse.equinox.event needs to be auto-started.
It was also necessary to tweak the start levels for this to work.
This was inspired by the platform's sdk.product file. See
http://git.eclipse.org/c/platform/eclipse.platform.releng.aggregator.git/tree/eclipse.platform.releng.tychoeclipsebuilder/sdk/sdk.product
Change-Id: Ie7e6cc7f5ba5d4fb35a7ae82d02d244bcfdba20e
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56457
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Bernd Hufmann [Wed, 2 Sep 2015 01:20:32 +0000 (21:20 -0400)] 
tmf: Bug 476390: Support SHIFT+left click to update the selection range
Change-Id: Ie11006e88c325c77b16d5d7bdb3765944b54e0f6
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/55050
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Bernd Hufmann [Thu, 3 Sep 2015 16:49:41 +0000 (12:49 -0400)]
tmf: Bug 477794: Persist trace group root during remote fetching
This will improve the import feature of LTTng Control the performance
because it doesn't scan the parent folder for the relevant traces of
a session.
Change-Id: I07739c7e229c2026cb8393a19061152a51678dde
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56224
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Marc-Andre Laperle [Tue, 22 Sep 2015 02:27:16 +0000 (22:27 -0400)]
tmf: Fix for visible middle scrollbar on Mac
On Windows we need to use a workaround to hide the vertical
scrollbar. By using that workaround on Mac, it actually makes the
vertical scrollbar visible all the time. Instead, we can use the
SWT.H_SCROLL | SWT.NO_SCROLL combination to only get the horizontal
scrollbar. This doesn't work on Windows but can still be used on Mac
and Linux (GTK).
Bug: 466870
Change-Id: If199708b41ed1b9467c6950d5fcf7ce7e461f1e1
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/50115
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Tested-by: Patrick Tasse <patrick.tasse@gmail.com>
Bernd Hufmann [Tue, 22 Sep 2015 14:07:30 +0000 (10:07 -0400)]
control: Fix dead store sonar warning in destroy session handler
Change-Id: I6fdc75a373cae524f45cd5d63fb1f724286f08fc
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56431
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Hudson CI
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Bernd Hufmann [Mon, 21 Sep 2015 16:50:05 +0000 (12:50 -0400)]
control: Don't hold a lock when in dialogs and add null annotations
Change-Id: I45d2e1d45eceeefd4c39ffed06a8f674c91d3e78
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56368
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Hudson CI
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Bernd Hufmann [Mon, 21 Sep 2015 11:16:36 +0000 (07:16 -0400)]
control: Bug 477630: Fix concurrency issue in destroy session handler
Use a lock to protect the list of sessions from concurrent access and
modification.
Change-Id: Ie671910b31b226b202e3ac70e02e32ca8fb9fd01
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56333
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Wed, 17 Sep 2014 20:48:23 +0000 (16:48 -0400)]
tmf: add import/export package swtbot tests
Change-Id: Ie4da7c0cd4b9dce58b770f776192d42306f3a291
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/37772
Reviewed-by: Hudson CI
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Tested-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Matthew Khouzam [Wed, 16 Sep 2015 20:49:37 +0000 (16:49 -0400)]
GDB: No longer synchronize on a string.
This is dangerous for two reasons:
1- if the string value changes, the object changes too, so we can
corrupt data.
2- if the string is internalized, it can cause unwanted delays with
other objects synchronizing coincidentally on the same string.
Change-Id: Ibfeb3702ef82424b2fca86956f8227b52ae8b945
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56133
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Tested-by: Patrick Tasse <patrick.tasse@gmail.com>
Alexandre Montplaisir [Fri, 11 Sep 2015 16:56:24 +0000 (12:56 -0400)]
analysis: Correctly track the state of scheduled-out processes
On a sched_switch, the state of a the process going out should not
always be WAIT_BLOCKED (the yellow state in the CFV). If the prev_state
field indicates 0 (TASK_RUNNING), it means the process was scheduled
out due to preemption, not to being blocked on something. In those
cases, we should put the process into the WAIT_FOR_CPU state.
Bug: 477052
Change-Id: Ic181811c11029d6374a5e65149a103febd25b5ea
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/55770
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Fri, 31 Jul 2015 23:48:02 +0000 (19:48 -0400)]
ctf: fix output writer segment intersection algorithm
Let's say you have two segments, A and B, they have the points
A0, A1, B0, B1. to see if there is intersection between the two
segments in one dimension, here are the exhaustive possibilities.
A----A
B--B 1
B--B 2
B--B 3
B--B 4
B--B 5
B----------B 6
we need to only reject 1 and 5
In 1, B0 && B1 < A0
In 5, B0 && B1 > A1
Now B0 < B1 so we can simplify to
B1 < A0 || B0 > A1
Applying demorgan we get
!(B1 >= A0 && B0 <= A1)
If B is the entry, B0 = entry.timestampStart() and B1 = entry.timestampEnd()
The intersection is the not(!) of the formula above.
entry.timestampEnd >= startTime && entry.timestampStart <= endTime
The patch changes the intersection algorith to that.
Change-Id: Ie3205a3c7da970daf94c0cb93cb336fe0851f039
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53001
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Thu, 27 Aug 2015 02:58:57 +0000 (22:58 -0400)]
ctf: Make CtfStreamInputReader NonNullByDefault
Change-Id: Idb221d898e7b734fddcacc31e3868f5f43a0c01a
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54647
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Bernd Hufmann [Wed, 16 Sep 2015 17:08:57 +0000 (13:08 -0400)]
tmf: Fix rounding error in seekEvent by ratio for TextTrace class
A similar fix was merged for the CustomTxtTrace (see 391277) but has
been never transferred to the TextTrace class.
Change-Id: I94267e47e4cf2b080b5a396ea3a98313c01a43bb
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56112
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Tested-by: Patrick Tasse <patrick.tasse@gmail.com>
Patrick Tasse [Tue, 15 Sep 2015 16:46:55 +0000 (12:46 -0400)]
analysis: Enable filter dialog in Resources view
Change-Id: I5c50ed03a072c4ea24aed839d765890c3773a6bb
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/56002
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Patrick Tasse [Tue, 15 Sep 2015 16:44:43 +0000 (12:44 -0400)]
tmf: Split time graph filter dialog from TimeGraphCombo
This allows the filter dialog to be used with a TimeGraphViewer.
Change-Id: I78323a0848ced5139fc96b49973a19998e564cbf
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/56001
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Patrick Tasse [Tue, 15 Sep 2015 16:42:23 +0000 (12:42 -0400)]
tmf: Fix packing of columns in TimeGraphFilterDialog
The packing should be done after the elements have been expanded.
Change-Id: I7e4351056391ffb76734ea3bae034dbb0982df8f
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/56000
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Mon, 13 Jul 2015 14:16:12 +0000 (10:16 -0400)]
ctf: Remove useless alignment check
Now that default alignment is 1, statements like "alignment =
((size % DEFAULT_ALIGNMENT) == 0) ? 1 : DEFAULT_ALIGNMENT;" make no sense.
A- the % 1 will always return 0 and B- the return values are either 1 or 1.
Change-Id: I30613fba9cb586b884675c55544727fea0efea89
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/51853
Reviewed-by: Hudson CI
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Tested-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Matthew Khouzam [Thu, 6 Aug 2015 15:04:22 +0000 (11:04 -0400)]
swtbot: fix npe in import tests
The trace helper getEvent(rank) no longer disposes the trace.
Change-Id: I6ee7753c4909c1eba231f0402eda4737d8c167ea
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53340
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Geneviève Bastien [Tue, 15 Sep 2015 18:53:06 +0000 (14:53 -0400)] 
ss: update some classes to current code standard
* Name fields fSomething
* Remove this. before field names
Change-Id: I49028e34a496e16aa5fa89be01fad7ef906d4fc0
Signed-off-by: Geneviève Bastien <gbastien+lttng@versatic.net>
Reviewed-on: https://git.eclipse.org/r/52754
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Hudson CI
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Patrick Tasse [Tue, 15 Sep 2015 15:27:07 +0000 (11:27 -0400)]
tmf: Fix incorrect TimeGraphFilterDialog result when dialog shell closed
The result is set to null when the Cancel button is pressed. But when
the dialog shell is closed by pressing the [x] widget, the result
remains set to the value set the last time that the dialog was closed
with a button. The first time that the dialog is opened, the result is
initialized to an empty array, so closing the shell filters all items.
Change-Id: I82225608e54a233edc9fa7ee66dd826c55d50c52
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55999
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Tue, 15 Sep 2015 18:55:47 +0000 (14:55 -0400)]
ctf.core.test: add trace to .gitignore
Change-Id: I0f52a1a266f03f15331e9dc9c8ec5ab04e485d73
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/56020
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Tue, 15 Sep 2015 18:33:53 +0000 (14:33 -0400)]
ctf: clean up EventDeclaration#equals
Use NonNullUtils.equalsNullable() to make code more legible.
Change-Id: I36ccfce40e41c71907483db357906fc96f3d4d42
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/48038
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-by: Hudson CI
Patrick Tasse [Tue, 15 Sep 2015 15:01:41 +0000 (11:01 -0400)]
tmf: Fix SWTException in FilteredCheckboxTree
The tree can be disposed by the time the refresh job is done.
Change-Id: Ic1f4761940dc2866762601945c1b93b3836db330
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55998
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Reviewed-by: Hudson CI
Patrick Tasse [Fri, 11 Sep 2015 14:50:17 +0000 (10:50 -0400)]
tmf: Persist filters of opened traces in AbstractTimeGraphView
The filters of each opened trace are stored in a map and restored when
switching from one trace to another.
Change-Id: I101961009eaf9123aca0995cc1e7ea862f13d262
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55749
Reviewed-by: Hudson CI
Patrick Tasse [Fri, 11 Sep 2015 22:33:12 +0000 (18:33 -0400)]
tmf: Fix TimeGraphCombo item height synchronization issues
- Remove the caching of item height data in the tree item widget that is
meant to prevent repeatedly setting the time graph item height.
This only saved a hash map lookup in the time graph, but required two
linear searches in the TreeItem widget keyed data. It also could prevent
the height alignment to be performed if the time graph changed its item
heights without the time graph combo knowing about it. For example, all
time graph item heights are reset on refresh.
- Set the input on the time graph before setting the global item height.
Getting the global item height involves refreshing the list of visible
expanded elements, which also synchronizes their expanded state. If the
time graph input is not set yet, this causes all tree items to be
unnecessary (but temporarily) collapsed. As as counterpart to changing
the order, setting the global item height on the time graph will update
the item height of every existing time graph item. This only happens
once (when the input is set).
Change-Id: Id3519cbf8f9ce283583ac75af9c4f3ec7174b24a
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55786
Reviewed-by: Hudson CI
Francis Giraldeau [Tue, 30 Jun 2015 02:15:03 +0000 (22:15 -0400)] 
Analysis: Add unit tests for dependency graph builder module
Change-Id: I214e1539b93705789bbb04e007ad893383d5cbbd
Signed-off-by: Geneviève Bastien <gbastien+lttng@versatic.net>
Signed-off-by: Francis Giraldeau <francis.giraldeau@gmail.com>
Reviewed-on: https://git.eclipse.org/r/51080
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Francis Giraldeau [Tue, 30 Jun 2015 01:50:53 +0000 (21:50 -0400)]
Analysis: Add the dependency graph builder module
Change-Id: Icca0b33d07b14eee7009a0abd8fd064da061eed0
Signed-off-by: Geneviève Bastien <gbastien+lttng@versatic.net>
Signed-off-by: Francis Giraldeau <francis.giraldeau@gmail.com>
Reviewed-on: https://git.eclipse.org/r/51079
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Francis Giraldeau [Fri, 11 Sep 2015 19:44:36 +0000 (15:44 -0400)]
analysis: improve graph statistics scalability
The TmfGraphStatistics was keeping all duration values. Individual
values are not required to display summary statistics. Instead, this
patch updates the sum per IGraphWorker. Also, we introduce the fTotal
field and remove the special placeholder String key in the map, allowing
to rectify the keys type from Object to IGraphWorker.
* update v2:
* synchronize only on fWorkerStats
* make only one map lookup to update worker sum
Change-Id: I202945bdad6a0cb759b2fc4cbb791b3afef7c3cc
Signed-off-by: Francis Giraldeau <francis.giraldeau@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55780
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Bernd Hufmann [Fri, 11 Sep 2015 18:57:24 +0000 (14:57 -0400)]
tmf: Update the initial item height calculation in Linux
In Linux the item height is calculated by the difference of the y
positions of 2 first consecutive tree items. When the 2 tree items
are available a paint listener is added to the tree and in the paint
method the height is calculated using the 2 stored items. The
paint listener is needed to make sure that the items are painted
and the y values are correct. However, between the time the items are
retrieved and the paint listener is executed the first item could have
been expanded and the y position second item is changed so that the
calculated height is much bigger. This patch fixes it by retrieving
the items in the paint listener.
Up to now this condition happens very, very rarely. However, with
patch https://git.eclipse.org/r/#/c/55749 this will happen consistently
Change-Id: I528c9baa5602cf8897ff84dc98cf71be2fb1fb91
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/55779
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Tested-by: Patrick Tasse <patrick.tasse@gmail.com>
Patrick Tasse [Thu, 10 Sep 2015 15:12:03 +0000 (11:12 -0400)]
tmf: Clamp single time selection to time graph bounds
When the selection is updated to a single time, it should be clamped to
the current bounds. This will make it consistent with the handling of a
selection range update.
Additionally, when the selection range is updated with possible
notification, the check to see if the selection range has changed is now
made by comparing the clamped values, not the received values. Although,
in practice, since this update is triggered internally, it should not be
received with out-of-bounds values.
Change-Id: Iee93f7ef401fa6efc087aa03687de51f7f994590
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55671
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Patrick Tasse [Thu, 10 Sep 2015 15:04:28 +0000 (11:04 -0400)]
tmf: Avoid zoom on selection range update in AbstractTimeGraphView
When the selection range is updated externally, the selection is set on
the time graph widget with the 'ensure visible' flag set to true. This
may or may not cause the window range to be updated.
The AbstractTimeGraphView unconditionally starts a zoom thread in case
that the window range has changed. It should start it only if the range
has changed.
However since the fix to Bug 476148, if the selection update causes the
window range to change, the AbstractTimeGraphView will always be
notified through the ITimeGraphRangeListener callback, and this will
start a zoom thread. So it is now unnecessary to worry about starting a
zoom thread on handling of the TmfSelectionRangeUpdatedSignal.
Additionally, in the handling of this signal, the call to method
synchingToTime() is changed to use the time graph's selection time, not
the one received in the signal, as the time graph may clamp the
selection time according to the current time bounds.
Change-Id: I9e5d1f69107f56d4940fb2a952b41cb150832635
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55670
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexandre Montplaisir [Fri, 4 Sep 2015 21:44:32 +0000 (17:44 -0400)]
tmf: Merge TmfEventProvider#getType() and ITmfTrace#getEventType()
TmfTrace has access to both getType() and getEventType(), which
come from two separate super class/interface, but do the exact
same thing.
We can simplify this by only declaring it once, in ITmfEventProvider
(which ITmfTrace extends).
Change-Id: I8a605da5bfd28548d36cd4ce18a1e94d48698176
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/55344
Reviewed-by: Hudson CI
Geneviève Bastien [Thu, 9 Oct 2014 21:11:45 +0000 (17:11 -0400)]
analysis: Fix the state of processes in the LTTng Kernel Analysis Module
A process with prev_state=1024 was shown as blocked, while it is actually
ready to be run on CPU, so that state must be ignored.
Change-Id: I0f2e7acd05e1592e5a91d898869a9a82ad66ead1
Signed-off-by: Geneviève Bastien <gbastien+lttng@versatic.net>
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/35492
Reviewed-by: Hudson CI
Marc-Andre Laperle [Mon, 7 Sep 2015 19:15:58 +0000 (15:15 -0400)]
ctf: Fix unreliable handling of invalid location
This could lead to bad seeks when opening experiments (bug 476816).
In experiments, it is often the case (expected) that a seek to a
location will result to an invalid location for one of the traces in
the experiment.
There are issues with the handling of seeking to an invalid
location (CtfLocation.INVALID_LOCATION)
1. In one spot, == is used instead of .equals which means that a
location read from the index is not considered invalid, but rather,
it seeks to an offset of 0.
2. Even if the location is properly considered invalid (== fixed),
the code relies on the "last packet end time" to seek past the end of
the trace which then make the context location become invalid. This
works correctly when the trace was fully read the first time but the
"last packet end time" is not necessarily the end of the trace when
it is opened a second time.
Instead, we can explicitly check for CtfLocation.INVALID_LOCATION
in CtfTmfContext.setLocation and CtfIterator.seek to make sure no
bad seeks occur.
Bug: 476816
Change-Id: I4c74259ae3b67fb22cae302f2d5f09f1adda51f0
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/55429
Reviewed-by: Hudson CI
Patrick Tasse [Wed, 9 Sep 2015 22:09:52 +0000 (18:09 -0400)]
tmf: Bug 477012: Time graph fails to send window range notification
Change-Id: I29b57b77db71b78deb0161e57c24135f3af60191
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55596
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Patrick Tasse [Wed, 9 Sep 2015 21:38:14 +0000 (17:38 -0400)]
tmf: Bug 477009: AbstractTimeGraphView sets wrong bounds
When no supported trace is selected, set the defaults bounds on the time
graph widget to [SWT.DEFAULT, SWT.DEFAULT] instead of [+INF, -INF].
Change-Id: I814fb72886a25be5d2fe935f5f0e37dc207686f5
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55593
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexandre Montplaisir [Tue, 7 Jul 2015 20:00:17 +0000 (16:00 -0400)]
btf: Remove AutoCloseable from BtfTrace
Despite many great advantages, AutoCloseable does not work well
with subclasses of ITmfTrace, because it would imply that such
objects are not exposed publicly by other objects tracking them.
This is not the case in TMF, as many objects expose ITmfTrace's
through getters. Using instead TmfTrace's standard dispose()
method is more appropriate in subclasses.
Change-Id: I1d342d15dba7f8f6e75985f8aee6ed159838d424
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51540
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Bernd Hufmann [Wed, 2 Sep 2015 19:11:57 +0000 (15:11 -0400)]
tmf: Bug 476551: Omit empty traces in standard import and remote fetch
Change-Id: I1a0307f98f479c28300ca4c42f4c47d4f1fdfc45
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/55188
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Tested-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexandre Montplaisir [Tue, 7 Jul 2015 20:29:01 +0000 (16:29 -0400)]
ctf: Remove AutoCloseable from CtfTmfTrace
Implementations of TmfTrace should not be AutoCloseable,
since they are exposed by other objects through getters.
Change-Id: Ibdb9b056802262cb3e0d67b6f5d3434f79c7face
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51542
Reviewed-by: Hudson CI
Alexandre Montplaisir [Tue, 7 Jul 2015 20:52:14 +0000 (16:52 -0400)]
pcap: Remove AutoCloseable from PcapTrace
Change-Id: Ide58410feef5b85bbe7c2c2a274b200bf614045b
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51543
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexandre Montplaisir [Tue, 7 Jul 2015 20:32:22 +0000 (16:32 -0400)]
lttng: Open and close trace only once in ImportAndReadKernelSmokeTest
Change-Id: Ieaae2b0871fcbb5ff7009b84cca678d83806d4ac
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51541
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Patrick Tasse [Fri, 4 Sep 2015 22:22:39 +0000 (18:22 -0400)]
doc: Update user guide for split aspect combo in filter viewer
Change-Id: I09158bbce40217e0211bbd4326836c52785243c0
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55345
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Patrick Tasse [Thu, 3 Sep 2015 19:10:21 +0000 (15:10 -0400)]
tmf: Bug 476481: NullTimeEvent should not affect time graph entry bounds
This fixes the process entry's Birth time in the Control Flow view.
Change-Id: If0cb2aa0c644d930720fd4e95b074e3edb748cd8
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55232
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Raphaël Beamonte [Thu, 3 Sep 2015 22:25:02 +0000 (18:25 -0400)] 
ctf: Add error logging when a closed iterator is selected for reattribution
In replaceRandomElement, a check has been added to create a new
iterator when encountering a closed one. This situation should
never happen, and so means that something is broken somewhere
else, e.g. that an ITmfContext hasn't been disposed properly.
This patch adds an error message log when that case is met to
allow further investigation and correction of the origin of
the problem.
Change-Id: Ib35f9906045f21042fbc9fea216a0e5da81221d3
Signed-off-by: Raphaël Beamonte <raphael.beamonte@gmail.com>
Reviewed-on: https://git.eclipse.org/r/55242
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Tested-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexandre Montplaisir [Wed, 2 Sep 2015 23:18:54 +0000 (19:18 -0400)]
ss: Add proper locking to the AttributeTree
Instead of using a Collections.synchronizedList, extend the locking
the whole object. This makes sure that data stays consistent in the
longer methods like getQuarkAndAdd()/getQuarkDontAdd().
Bug: 476474.
Change-Id: Ia57f77ad109dd8c0b90e1824ba04d3a57140c1b6
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/55156
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Hudson CI
Patrick Tasse [Mon, 31 Aug 2015 21:45:47 +0000 (17:45 -0400)]
tmf: Split aspect combo in filter viewer
Instead of a single aspect combo containing all trace types and their
respective event aspects, use a first combo for the trace type and a
second combo for the event aspects corresponding to the selected trace
type in the first combo.
Change-Id: I9e6189590d701ac8fea048a59e16d064491ccd01
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54941
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Marc-Andre Laperle [Mon, 31 Aug 2015 14:44:04 +0000 (10:44 -0400)]
rcp: Use rcp repository as source for creating the installations
With this option, the build implicitly verifies that it would also
be possible to install the product (and its features) from that
repository.
This means that it provides some basic sanity check for installing
Trace Compass features. At least, for the ones included in the RCP.
For example, it would have uncovered the missing segement plugin in
the tmf feature, fixed here: https://git.eclipse.org/r/#/c/54804/
See
https://eclipse.org/tycho/sitedocs/tycho-p2/tycho-p2-director-plugin/materialize-products-mojo.html#source
Change-Id: I54e8ab63f830cbdb712b6c5c88d2c344f38e4518
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54918
Reviewed-by: Hudson CI
Reviewed-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Tested-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Alexandre Montplaisir [Mon, 31 Aug 2015 22:51:02 +0000 (18:51 -0400)]
Build the RCP by default
Activate the "build-rcp" profile by default, and document how to
disable it to skip building the RCP.
Change-Id: I118db241eac28446121a4a6f5d1b71f1e5a7f1fb
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/54943
Reviewed-by: Hudson CI
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Tested-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Marc-Andre Laperle [Wed, 19 Aug 2015 20:01:16 +0000 (16:01 -0400)]
tmf: Add a more reliable way to get the active editor with SWTBot
StandardImportGzipTraceTest assumes that after double-clicking on the
trace, there will be an active editor right away. Instead, we can
have a method that waits until there is an active editor.
I've seen the failure:
WidgetNotFoundException: There is no active editor
here:
http://eclip.se/5l
Change-Id: I41dd5c7c4f30f789dabab813eb8bcbb81fdcb48e
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54154
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Patrick Tasse [Thu, 27 Aug 2015 21:07:00 +0000 (17:07 -0400)]
tmf: Bug 476148: Fix time range synchronization problem in time graph
Change-Id: Idafa2029da4c57d32f85414f3545401c13b91b11
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54816
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Marc-Andre Laperle [Mon, 31 Aug 2015 19:31:42 +0000 (15:31 -0400)]
tmf: Change TmfEventsTable to not use Display.getCursorLocation
This makes CallsiteEventsInTableTest (SWTBot) more stable because
Display.getCursorLocation is getting the cursor location straight
from the windowing system and therefore does not work properly with
SWTBot. Instead, we can use the last position of the MenuDetect
event to know the location.
Also update the targets sequence numbers to pick up required fix
from SWTBot, see SWTBot bug 474747.
Bug: 474435
Change-Id: I7e8f4377e93d78030ebe582e0b4a92666a81e966
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53515
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Marc-Andre Laperle [Sun, 30 Aug 2015 01:37:19 +0000 (21:37 -0400)]
tmf: Fix Negative seek offset in indexer
If a Btree is saved to disk without any checkpoints stored in it, the
root node is not serialized out and invalid data is read when
reopening the index.
This is because the dirty flag for a node is set to false initially.
This works correctly for all child nodes because they are always
modified at least once right after being allocated but this is not
true for the root node since it is always allocated regardless of
whether or not it will contain checkpoints (and other child nodes).
Setting the dirty flag to true initially fixes the issue because
this will ensure nodes are serialized-out at least once. Nodes that
are serialized-in are not dirty until modified again.
Bug: 475543
Change-Id: I62db56445951f40ce0f27af04dd6f0d927f0c2fc
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54841
Reviewed-by: Hudson CI
Marc-Andre Laperle [Mon, 31 Aug 2015 22:08:28 +0000 (18:08 -0400)]
tmf: Fix wrong context menu in ColumnHeaderMenuTest
This worked before because SWTBot didn't send the appropriate
MenuDetect when right-clicking on a table so the TmfEventsTable
still had the last context menu (the header's). Now that it does
send MenuDetect, it right-clicks in the middle of the table and the
context menu is not the header's context menu. The context menu call
should be done on the header instead.
Change-Id: I205d3f1de12d0b8b04100d956f5c1557aa10f57d
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54940
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-by: Hudson CI
Alexandre Montplaisir [Sun, 30 Aug 2015 22:16:01 +0000 (18:16 -0400)]
ctf: Smooth out some API breaks in CtfTmfEvent
- Keep getCPU() instead of getCpu(), not much value in changing this.
- Keep a @Deprecated getReference() that will call getChannel().
Change-Id: Ib5f603c926bc9db35423ba829d3696dc2dcfa60f
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/54857
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Thu, 27 Aug 2015 03:05:33 +0000 (23:05 -0400)]
ctf: add details to bitbuffer exception
Change-Id: I4b64f99112644dca381edb40275e4f3990686326
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54648
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Thu, 27 Aug 2015 02:38:06 +0000 (22:38 -0400)]
ctf: remove getParent() from CTFStreamInputReader
Bump the API to 2.0, remove a useless coupling method and field
Change-Id: I951236c3b401c00ae57e9080fb9e4fb47ecb73fb
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54646
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Thu, 27 Aug 2015 02:29:01 +0000 (22:29 -0400)]
ctf: make StreamInputPacketIndexEntry start at time 0
Change-Id: I77302be9b2f26b53bb4c4fddd8c5d9808625b470
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54645
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Bernd Hufmann [Fri, 28 Aug 2015 18:36:55 +0000 (14:36 -0400)]
tmf: add segment store plugin to tmf feature
... otherwise it won't be part of the update site and installations
would fail
Change-Id: I07fca482b139c6e4d440a47c916d61ff546bf671
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54804
Reviewed-by: Hudson CI
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Tested-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Matthew Khouzam [Mon, 24 Aug 2015 15:03:10 +0000 (11:03 -0400)]
lttng: make dead tasks return the state to null instead of blocked.
This will make the control flow view show a task finish instead of go
on and on and on. After this, PIDs can be more clearly re-used.
This patch updates the test cases as the behaviour of the state provider
changes.
Change-Id: Ib9fd24a868734b94cca9063c6753d88aa1185eb0
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54415
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Wed, 26 Aug 2015 20:17:39 +0000 (16:17 -0400)]
analysis.linux: add TASK_STATE_DEAD and javadoc to LinuxValues.java
TASK_STATE_DEAD is useful to check if a task is still runnable or not.
This will be useful to see if a task can run or not when it is scheduled
out.
I had to read the kernel code for a while to figure out what
some of the flags mean, this should help others save time.
Change-Id: If538ce6b928c40ccf985ea655c58ce2da268539d
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54630
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Reviewed-by: Hudson CI
Patrick Tasse [Fri, 28 Aug 2015 15:22:46 +0000 (11:22 -0400)]
tmf: Bug 476129: NullPointerException in GenerateTestValues.java
The preferences service is not initialized since GenerateTestValues does
not load the Eclipse platform. The default preferences should be used.
Change-Id: I8a033bc59e224e8b9aac79cf5e1e7f21d5b24b75
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54788
Reviewed-by: Hudson CI
Reviewed-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Tested-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Bernd Hufmann [Tue, 11 Aug 2015 17:51:27 +0000 (13:51 -0400)]
control: replace ImportDialog with TMF remote fetch wizard
Change-Id: I14ad7587e0e278f511d83c586b4cfd65306877f6
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53604
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexandre Montplaisir [Wed, 1 Jul 2015 00:04:39 +0000 (20:04 -0400)]
tmf: Add unit tests for TmfEventField using arrays as values
Change-Id: I021dec1e8ee7fafdd67b5a8ae1faa8343636de44
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51214
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Alexandre Montplaisir [Wed, 15 Jul 2015 23:56:55 +0000 (19:56 -0400)]
ctf: CtfTmfEvent improvements
- Rename getReference() to getChannel().
- Rename getCPU() to getCpu().
- Remove getID(). It was not used by anything, and as was mentioned
in its Javadoc, it's better to use the event name instead, so it
is not very useful to expose it.
- Add @NNBD, and related annotations and checks
- Add equals() and hashCode() methods.
Change-Id: I83eb42b01b98895bbdb63a8c6fdba7249f93ec06
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/52121
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Alexandre Montplaisir [Tue, 30 Jun 2015 21:18:10 +0000 (17:18 -0400)]
tmf: Use getter methods instead of fields in TmfEvent#equals()
Since the various getter methods in TmfEvent are not final,
which means that subclasses can (and do) extend them, the
method calls should be used instead of the object's fields
to correctly compare two such objects.
Refs bug #471733.
Change-Id: I2f4a2b8f455cfecd2b2a11f480092cd751eb50d3
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51139
Reviewed-by: Hudson CI
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Alexandre Montplaisir [Wed, 26 Aug 2015 16:02:10 +0000 (12:02 -0400)]
tmf: Add a timeout to TmfTraceManagerTest
Change-Id: I93666b60a88cd83afc3502ef9b77b5e444507a4c
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/54618
Reviewed-by: Hudson CI
Matthew Khouzam [Fri, 21 Aug 2015 21:03:38 +0000 (17:03 -0400)]
ctf: make StreamInputPacketIndex more thread safe
This patch makes append synchronized. The SIPI can only grow and elements,
once in the index, cannot be changed. Therefore synchronization on reading
does not seem essential. If the same packet is written twice, the second
time, it will fail gracefully. The packet will still be there. This
should help "CtfTmfContextTest.testTooManyTraces".
Change-Id: Ic4dfcdda71fc1ebb58039066dab836e67458506f
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54344
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
France Lapointe Nguyen [Fri, 21 Aug 2015 22:06:46 +0000 (18:06 -0400)] 
analysis: introduce latency table view
Change-Id: Icdad089daf432306146cd390bf5694a95dc0b5f8
Signed-off-by: France Lapointe Nguyen <francelap@gmail.com>
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/52255
Reviewed-by: Hudson CI
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Alexandre Montplaisir [Wed, 19 Aug 2015 22:37:20 +0000 (18:37 -0400)]
analysis: Export a package that was missing
Change-Id: I697dd93d410692cb97f820ce5ab482aa4ba96d82
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/54176
Reviewed-by: Hudson CI
France Lapointe Nguyen [Fri, 21 Aug 2015 21:51:44 +0000 (17:51 -0400)]
tmf: support sorting lazy content provider in TmfSimpleTableViewer
Change-Id: I9e45b78473dac2600f7b4a2b3ea07b14d5b06615
Signed-off-by: France Lapointe Nguyen <francelap@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54352
Reviewed-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-by: Hudson CI
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Alexis Cabana-Loriaux [Fri, 21 Aug 2015 20:57:16 +0000 (16:57 -0400)] 
piecharts: add tree selection when a piechart slice is selected
Change-Id: I2450a060cf72a2d2661b9196d87c693326cf7be5
Signed-off-by: Alexis Cabana-Loriaux <alex021994@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54162
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Alexis Cabana-Loriaux [Thu, 20 Aug 2015 21:59:41 +0000 (17:59 -0400)]
tmf: make piecharts not visible when only one event type
Change-Id: I165811275f378aeecab7b4462b257b0c94c3d990
Signed-off-by: Alexis Cabana-Loriaux <alex021994@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54268
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Patrick Tasse [Tue, 11 Aug 2015 20:33:50 +0000 (16:33 -0400)]
tmf: Add support for grid lines in time graph
This adds support for optional grid lines in the time graph control. The
default is that grid lines are enabled. The grid line color can be set.
Change-Id: Ia52371e05d83a951e2e5babe0fb36c62f08c6d13
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54039
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Patrick Tasse [Wed, 12 Aug 2015 19:02:40 +0000 (15:02 -0400)]
tmf: Avoid redrawing background for every time graph entry.
The whole background is already filled for the bounds of the time graph.
It is unnecessary to draw it again for every time graph entry. For the
selected entry and entries with no time events, the background with
different color and/or bounds needs to overwrite the default. The drag
zoom range does not need to be painted for every time graph entry, it is
already painted over with transparency after all items have been drawn.
These backgrounds are painted in one step before any of the item states
are drawn. This will allow to paint elements that should be below the
time events (such as gridlines) over the time graph background without
them being overwritten by the unnecessary redrawing of the background.
The computation of item, name and state rectangle bounds is optimized.
The state rectangle height computation is changed to ensure that the
default margin height of 3 pixels does not make it too small.
Change-Id: I2ef87fb3ab2963f5dfeca3f3d1c7d55618dead23
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54038
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexandre Montplaisir [Thu, 13 Aug 2015 20:05:58 +0000 (16:05 -0400)]
Document the -Dtarget-platform option in the README
Change-Id: I3141344e1a71a289a0e5b530ec6521507e3b3460
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/53738
Marc-Andre Laperle [Mon, 17 Aug 2015 00:55:18 +0000 (20:55 -0400)]
tmf: Make tracing log to stdout, like other Eclipse tracing options
When enabling tracing in Eclipse, things are usually logged to stdout with
some plugins having the option to log to file as well. I think to be more
consistent, Trace Compass should log to stdout. It's also a bit more
convenient if you don't need to keep the output.
Change-Id: I57517ab43c007ba1f5f1f249ab114c75cdd17243
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53847
Reviewed-by: Hudson CI
Patrick Tasse [Wed, 19 Aug 2015 17:34:09 +0000 (13:34 -0400)]
ss: Bug 475300: Fix inconsistent TreeMapStore due to value comparators
Add Comparable to the ISegment interface for segments to specify
a natural ordering. This ordering will then be used by the tree map
to "break ties" when segments have the same start and end times.
Also removed the getElementAtIndex() method. It was not doing what
the Javadoc said it did, and it does not seem to be useful for the
analysis as it is.
Change-Id: Ic65f50b12f6e94c59d4a60f0e96d9c78ee04b741
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/54146
Reviewed-by: Hudson CI
Marc-Andre Laperle [Mon, 17 Aug 2015 03:37:19 +0000 (23:37 -0400)]
tmf: Support resuming the trace index
This allows resuming the indexing job after a trace was closed while the
indexing was not finished.
This also allows to index new content that was added to a trace, even if the
content was added after closing the trace and the indexer was finished.
Change-Id: I58296c0114a9f495605836f8c6df6dfceb424654
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53851
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Reviewed-by: Hudson CI
Alexis Cabana-Loriaux [Wed, 19 Aug 2015 20:56:32 +0000 (16:56 -0400)]
tmf: make piecharts not visible when no trace opened
Change-Id: Id59ec9bb8a6c861b0de98cb4a5072f075add85b0
Signed-off-by: Alexis Cabana-Loriaux <alex021994@gmail.com>
Reviewed-on: https://git.eclipse.org/r/54171
Reviewed-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Bernd Hufmann [Mon, 10 Aug 2015 14:24:48 +0000 (10:24 -0400)]
tmf: Bug 473280: Internalize TMF Statistics UI API
Change-Id: I0031b2fcb45dd52144d8a3feb3cd348d74467c5d
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53491
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-by: Hudson CI
Matthew Khouzam [Thu, 6 Aug 2015 13:59:15 +0000 (09:59 -0400)]
ctf: Lose reference to StreamInputReader in EventDefinition
StreamInputReader is AutoCloseable, it is better to not have a
reference to it in the event.
Change the constructor to require the CPU number instead of the
whole StreamInputReader, as it is the only thing that the
EventDefinition uses.
While we are changing most calls, we can merge both versions of
the constructor into one, since they only differ by one @Nullable
parameter.
Change-Id: Ibeae8e7ff7e4f78ab1f11a24d61c739706bfc881
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/53327
Reviewed-by: Hudson CI
Matthew Khouzam [Thu, 20 Aug 2015 04:36:22 +0000 (00:36 -0400)]
ctf: accelerate eventheaderdefinition timestamp parsing
Change-Id: I9d244f317c9974c4aea923f58852a26d7df6962b
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54184
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Matthew Khouzam [Tue, 18 Aug 2015 00:45:03 +0000 (20:45 -0400)]
ctf: update api to 2.0 baseline
Change-Id: Id29a6346da26885f918d55136abd21ab05268b8b
Signed-off-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53923
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexis Cabana-Loriaux [Fri, 31 Jul 2015 22:07:24 +0000 (18:07 -0400)]
tmf: extract UpdateJob class and introduce TmfPieChartViewer
Change-Id: I0371ed878bce9e3121eed98141248dac80dec95d
Signed-off-by: Alexis Cabana-Loriaux <alex021994@gmail.com>
Reviewed-on: https://git.eclipse.org/r/53160
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Bernd Hufmann [Wed, 19 Aug 2015 11:53:39 +0000 (07:53 -0400)]
releng: update feature versions to 2.0
Change-Id: I891313eae9fb1428d6007f08b46939edf33bcdca
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54091
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Marc-Andre Laperle [Mon, 17 Aug 2015 02:49:25 +0000 (22:49 -0400)]
tmf: Add grouping separator in indexer job
This makes it easier to read big numbers and track the progress of the job.
Change-Id: I2fc4329b378e71c78a738200b5c0886e183118c6
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53849
Reviewed-by: Hudson CI
Patrick Tasse [Thu, 6 Aug 2015 17:51:08 +0000 (13:51 -0400)]
ctf: Update CtfTmfContextTest
- Dispose the trace after the test
- Remove unused array
- Avoid extra expected value due to math rounding
Change-Id: I46e8fdff95dd3e39d12b786d5efc4560b46fefe5
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/53357
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Tested-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Marc-Andre Laperle [Fri, 14 Aug 2015 18:27:54 +0000 (14:27 -0400)]
[CTF] Events wrongly parsed as Lost Events
In CTFStreamInputPacketReader, setCurrentPacket can be called multiple times
without necessarily calling readNextEvent, which means that the cached value
for fHasLost can be wrong the next time readNextEvent is called.
Bug: 475007
Change-Id: I36c22c22c144cb8307403f93c76c996c60ff52aa
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53811
Reviewed-by: Hudson CI
Reviewed-by: Matthew Khouzam <matthew.khouzam@ericsson.com>
Marc-Andre Laperle [Tue, 18 Aug 2015 20:59:17 +0000 (16:59 -0400)]
Replace a few Boolean.valueOf(foo).booleanValue() with Boolean.parseBoolean
Change-Id: Ia45aa4f137f051f7a4cb650fddcb452249d7d874
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/54033
Reviewed-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Tested-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-by: Hudson CI
Marc-Andre Laperle [Mon, 17 Aug 2015 02:38:42 +0000 (22:38 -0400)]
tmf: Add indexer tracing option and add some trace points
Change-Id: I2a29558b197c0e25b068fdb3139539643d80e2a5
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53848
Reviewed-by: Hudson CI
Reviewed-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Patrick Tasse [Tue, 11 Aug 2015 18:10:52 +0000 (14:10 -0400)]
tmf: Remove unused filter methods in time graph control
The methods getTraceFilter() and getFilteredOut() were not used and
their implementation did not return a valid result.
Change-Id: Id1d330acab4e36a80259bfead3eccbeb3ffce82e
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/53587
Reviewed-by: Hudson CI
Alexandre Montplaisir [Fri, 10 Jul 2015 19:45:54 +0000 (15:45 -0400)]
releng: Add a Maven profile to run a custom test suite in alltests
Change-Id: I8484e5de675ad0b469182e3e2e1aeacdd380170a
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51777
Reviewed-by: Hudson CI
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Tested-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Marc-Andre Laperle [Mon, 17 Aug 2015 18:58:40 +0000 (14:58 -0400)]
tmf: Make StandardImportGzipTraceTest more robust
In a build, I got the failure:
"IllegalArgumentException: The row number (2) is more than the
number of rows(1) in the table"
http://eclip.se/5i
I think it's a case of the assertion hapenning before the table
was completely filled. Adding a few waitUntil should fix it.
Change-Id: I89b15a29bd0ae9f64f213fa0ffa715566127650a
Signed-off-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53915
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Patrick Tasse [Tue, 11 Aug 2015 14:34:22 +0000 (10:34 -0400)]
tmf: Bug 474710: Time graph combo expanded state is not synchronized
When a time graph combo is refreshed, newly discovered items are
expanded to the auto-expand level in the time graph viewer, and the tree
viewer item expanded states are synchronized with those of the time
graph viewer. The expanded state of existing items is no longer affected
by the call to refresh().
Change-Id: I709681ef488e7d22991d5b7384c667532afd3410
Signed-off-by: Patrick Tasse <patrick.tasse@gmail.com>
Reviewed-on: https://git.eclipse.org/r/53558
Reviewed-by: Hudson CI
Reviewed-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Tested-by: Bernd Hufmann <bernd.hufmann@ericsson.com>
Alexandre Montplaisir [Wed, 15 Jul 2015 15:59:51 +0000 (11:59 -0400)]
tmf: Add a timeout to AnalysisModuleTest
Change-Id: Ieb5c436b53760002ce3e35769bda2641ade92a4b
Signed-off-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
Reviewed-on: https://git.eclipse.org/r/51994
Reviewed-by: Hudson CI
Bernd Hufmann [Thu, 13 Aug 2015 13:48:07 +0000 (09:48 -0400)]
build: add minimum version to guava references in tmf.remote.core
Change-Id: If901686b2b9358160a5237914533fc4985ac5168
Signed-off-by: Bernd Hufmann <Bernd.Hufmann@ericsson.com>
Reviewed-on: https://git.eclipse.org/r/53710
Reviewed-by: Hudson CI
Reviewed-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Tested-by: Marc-Andre Laperle <marc-andre.laperle@ericsson.com>
Reviewed-by: Alexandre Montplaisir <alexmonthy@voxpopuli.im>
This page took 0.047729 seconds and 5 git commands to generate.
projects / deliverable / tracecompass.git / log
