| Commit message (Collapse) | Author | Age |
|
|
|
|
|
|
| |
Have the versions of the tools in the Makefile and automate the updating
of them when the numbers in the Makefile change.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
|
|
| |
The graph had a little 'hello' test to prove that the proper
pointer was being sent to the destroy function.
Remove it.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
| |
Have the gui built with 'make gui'. This allows for those that don't
care about the gui to just build the command line (by default).
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
| |
Remove the debug printing, or at least hide it.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
|
|
|
|
| |
When an event is by itself, information is displayed about it.
But it was displaying the wrong information.
This patch adds a separate variable to keep track of what should
be displayed and removes the confusion with trying to use other
variables.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
| |
The event info box was showing the wrong task if the pointer was before
the first record or after the last.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
| |
The event info box was getting confused by sched_switches, and showing
the wrong task at the given time line.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
| |
If the first event in the graph is a sched_switch, then the color
for the block was showing the new task. Not the previous task.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
|
|
|
| |
It is difficult to zoom in from the beginning of the graph, because
you must move the pointer right to zoom. To ease zooming in at
the start of the graph, this patch now allows the user to select
on the CPU drawing, and move the pointer over the graph to make
the selection.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
|
| |
When the pointer leaves the drawing area, sometimes the event info box
would stay hanging around. This looks messy, so remove it when the
pointer leaves the drawing area.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
If the scrolled window is vertically adjusted so that the top CPU
bar is close to the top of the viewable area, the event info box
can be hidden.
Change the location of the event info box to go beneath the pointer
when near the top.
The location of the box is also changed to be to the left of the
pointer instead of the right. This is because when we make the flip
the box may be hidden by the cursor itself. Although this does not
protect against that happening when the pointer is near the top
left corner, since the box will be to the right of the cursor then
anyway.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
| |
For some reason we called the horizontal adjustment of the graph vadj.
Call it hadj.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
| |
Add a search feature to search for matches in the list rows.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
| |
This patch lets wake up events that are waking up a filtered
task still show in the list, even if the waker is not filtered.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|
|
|
|
|
|
|
|
|
| |
When a task is filtered, have the wake up of the filtered task
also displayed.
This patch also modifies the event info box to show the task being
woken up instead of the waker.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| | |
When 64 bit needs to process %llu, it needs to convert it
first to %lu and use a long. But the code to do this had a bug
in the memmove, where the sorce and dest were reversed.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The list of events for the event filters, now only expands the system
rows that are not full or empty. That is, if the system is checked
it is not expanded. If the system is not checked and has no events
checked, then it is not expanded.
But if the system is not checked, and there are events checked
(even all events) then the row is expanded.
This makes dealing with lots of events much easier.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| | |
Keep the current row selected (or at least near by)
when updating the list event filter.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| | |
Selecting a row use to give errors when there were no rows.
Just prevent a row from being selected when there are none.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Let the list and the graph have their own filters.
While analyzing the data, I found it nice to be able to limit the graph
but still see other events in the list.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| | |
Have the graph send callbacks to update the lists filters.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| | |
Pass the task filters for both hiding and filtering from the
graph to the list.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| | |
Need to copy the count of the hash, not just the data.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Add option to hide tasks. This acts as the opposite of filtering.
If a task is added to the hidden list, then it will not show up.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| | |
Use the event filter callback to also update the graph.
Signen-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|\| |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Have git ignore the data files used by trace-cmd. The default data
file is trace.dat, but anything ending in .dat should be ignored.
Also, ignore patches that are laying around in the directory.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| | |
Move some of the trace-view event filtering to be generic enough and
have the trace-graph use it.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Since the event and CPU dialogs may also be used by the trace-graph
make it generic and remove the dependency of the trace-view-store.
This also changes them to accept a callback function for when a response
is made.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| | |
Apparently we have to free this.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| | |
The pango layout that is created for determining size is never
freed. Free it!
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| | |
With recent changes to getting rows, the assertion to check the
row integrity was not updated, and would cause false failures.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
A printf format has %llu for reading guint64, which is necessary
on 32bit boxes. But unfortunately, on 64 bit boxes, guint64 is a long.
This makes it inconsistent for how to read it. Just typecast it to
long long and shut up the warning.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| | |
If the event filter system or all-events is deselected, have
it deselect all its siblings. This is useful if you only want
to see a few events.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|\ \
| | |
| | |
| | | |
git://git.kernel.org/pub/scm/linux/kernel/git/dvhart/trace-cmd into dvhart-temp
|
| | |
| | |
| | |
| | | |
Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Rather than provide an accessor function for each field of the row
record in the trace-view-store, simply return the row record and allow
the caller to access the fields directly.
Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The trace graph shows filtered tasks scheduling in, but it
filters out that task from scheduling out. This fixes that bug.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Fix the list to view the task scheduling in. That is, the record belongs
to the previous task, but we want to see a filtered task scheduling in.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This patch now implements the event filtering dialog for the trace-view (list)
only. This works with kernelshark too, but only the list is filtered.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The event dialog now implements check boxes that are used to select and
unselect the events.
Although the event filter dialog is still not hooked to anything, and
does not affect the list yet.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| | |
| | |
| | |
| | | |
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Set the scrollbars to automatic, so they are not shown when not
needed.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| | |
The tree-view-main.c added the tree view to the scrolled window with
gtk_scrolled_window_add_with_viewport, but tree-views have their own
viewport. Not only is this not the proper way, but it was causing
a hang when we opened either the CPU or events dialogs.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| | |
Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
LKML-Reference: <1262654682-20325-4-git-send-email-dvhltc@us.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Make the necessary trace-view-store methods non-static and place their
declarations in the header file.
Functions beginning with _ are considered private in python. The structs
beginning with _ caused their accessor functions to be generated as
private. Since they aren't used as structs anywhere in the C code,
rename them to something more accessible to python.
Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
LKML-Reference: <1262654682-20325-3-git-send-email-dvhltc@us.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
trace_view_get_row() encapsulates some of the bounds checking and
provides a Python friendly way to access the store rows.
Signed-off-by: Darren Hart <dvhltc@us.ibm.com>
LKML-Reference: <1262654682-20325-2-git-send-email-dvhltc@us.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If we do a make clean, still allow the dependencies of the programs
to create the library. That is:
$ make clean
$ make trace-view
Needs to build the libtracecmd.a file.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|
| |
| |
| |
| |
| |
| |
| | |
Include when a task is scheduled in as an event in the filtered
list.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
|