Age | Commit message (Collapse) | Author |
|
|
|
* siri/cdv/abort-tag/OTP-14053:
[crashdump_viewer] Display abort reason when truncated
|
|
|
|
* siri/cdv/faulty-column-vals/OTP-14064:
[crashdump_viewer] Show correct values for 'Object' and 'Memory'
|
|
* richcarl/default-compiler-flags/PR-1226/OTP-14071:
stdlib test suite: fix uses of export_all
diameter tests: Eliminate use of -compile(export_all)
asn1 test suite: Suppress warnings for -compile(export_all)
Remove left-over uses of -compile(export_all)
Make warn_export_all the default
warn_obsolete_guard is already default
|
|
If a crashdump is truncated due to size limit reached, a new 'abort'
tag with reason is added at the end of the crashdump. This reason is
now displayed along with the truncated-warning.
|
|
A bug caused the numer of buckets to be shown in the 'Objects' column,
and the number of objects to be shown in the 'Memory' column for ets
table in crashdump_viewer. This is now corrected.
|
|
|
|
deleted: priv/crashdump_viewer.tool
deleted: priv/crashdump_viewer/collapsd.gif
deleted: priv/crashdump_viewer/exploded.gif
|
|
This sets the maximum queue size for the IP trace driver which is used
when tracing to shell and/or <c>{local,File}</c>.
Also, change the default queue size set by dbg:trace_port/2 to 200.
|
|
When observing mnesia table on remote node and connection to that node
is lost, a loop of error messages occur if mnesia is not running on
the observer node. ERL-237
|
|
Caused a crash in 19, see ERL-171.
Data was never reset so the number of schedulers didn't match the
previous collected data.
|
|
* josevalim/supervisor-get-callback-module/PR-1000/OTP-13619:
Return callback module in supervisor format_status
|
|
The previous implementation of supervisor:get_callback_module/1
used sys:get_status/1 to get the supervisor inner state and
retrieve the callback module. Such implementation forbids any
other supervisor implementation that has an internal state
different than the #state{} record in supervisor.erl.
This patch allows supervisors to return the callback module
as part of the sys:get_status/1 data, no longer coupling the
callback module implementation with the inner #state{} record.
Notice we have kept the clause matching the previous
sys:get_status/1 reply for backwards compatibility purposes.
|
|
|
|
* siri/observer/improve-trace/OTP-13481:
[observer] Update user guide
[observer] Automatically add active node
[observer] Allow more trace flags on procs/ports from GUI
[runtime_tools] Don't trace the trace client port
[runtime_tools] Allow setting trace flag 'exiting' with dbg
[observer] Add more default match specs for messages
[observer] Add test of new Ports tab
[observer] Improve appearance in Trace tab
[observer] Allow multiple select in Ports tab
[observer] Make right click menu act on the "expected pid"
[observer] In Trace tab, show procs/ports for selected node only
[observer] Add right click menu in Table tab
[observer] Add menu option to set default MS for send/receive
[observer] Add tracing of ports
[ttb] Allow setting trace flags on ports
[observer] Set correct parent in Label dialog
[observer] Add Ports tab in GUI
[observer] Add functionality in GUI for trace pattern on messages
[ttb] Set trace patterns on messages
Conflicts:
lib/observer/src/observer_wx.erl
|
|
When setting trace flags on 'new_processes' or 'new_ports', the
currently active node is now automatically added if no other node is
traced.
|
|
|
|
|
|
* Only allow single selection of modules
* Add right click menu to remove module
* Allow multiple selection of nodes, and show procs/ports for ALL
selected nodes
* On right click, only show menu items that make sense for the current
content and selection
* Add tooltips with help text
|
|
This is helpful when selecting ports to be traced.
|
|
If multiple processes were selcted, "Process Info" and "Kill Process"
menu choices would pick the first pid in the selection list, i.e. the
last selected process. This is now changed so the process under the
mouse pointer is used instead.
|
|
Earlier, all traced processes and ports would be shown, independent of
selected nodes in node view. This is now corrected.
|
|
|
|
|
|
|
|
|
|
The dialog for setting label on match specs in observer had faulty
parent. This is now corrected.
|
|
|
|
|
|
Functions ttb:tpe/2 and ttb:ctpe/1 are added.
|
|
Try to make observer stop nicer, i.e. without async crash reports
|
|
Tests on slow vnc server times out.
|
|
Looks better IMHO
|
|
And cut the line after 80 chars
|
|
Add extra space for ":" which is not accounted for size calculation.
|
|
Causes error reports on wxWidgets-3.1 backend
|
|
Could crash if interval was updated before node was set.
|
|
Bug introduced in bde2d01506a63, does work on linux but not on
the other OS's.
|
|
Memory table was flickering badly on Windows, caused by constant
refresh sent. Solve by letting the graphs have it's own parent.
wxWindow:setDoubleBuffer could help but causes an empty table
with the current version of wxWidgets backend.
|
|
Used to be misused wxTextCtrl's but looks better with wxStaticText, at
least on gtk we miss the posssbility to changes colours when hovering
over them.
|
|
|
|
OTP-13500
|
|
|
|
* dgud/observer/perf-freq:
observer: Optimize drawing of graphs
observer: Make it possible to change the graphs Xaxis and update freq
|
|
|
|
By allowing changes to the number of minutes displayed and update
frequency, we need to optimize the drawing of the graphs as it can
no longer recalculate everything in each frame drawn.
Only recalculate the changed entries, takes more memory but far less
cpu usage.
While updating the gui, increase the frame-rate a bit so it updates
smoother and decrease pen size for graphs to 1 pixel as it looks better
according to an office voting.
|
|
|
|
|
|
Crashed when term was displayed
|