https://bugs.kde.org/show_bug.cgi?id=396290
roger.li...@cedalo.com changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #114245|0 |1
is obsolete| |
--- Comment #19 from roger.li...@cedalo.com ---
Created attachment 114553
--> https://bugs.kde.org/attachment.cgi?id=114553&action=edit
Updated patch with better control.
This is a work-in-progress update in case there are comments while I'm working
on the documentation and tests.
* Renamed to failgrind
* Added failgrind.h with macros for turning on, turning off, and toggling
allocation failures, and clearing the in-memory call stack list.
* Added monitor commands to do the same, plus write the current in-memory call
stack list to a file, and print or zero stats.
* Split --callstack-file into --callstack-input and --callstack-output to allow
these to be independent
* Added --alloc-threshold-high, --allow-threshold-low, and
--alloc-threshold-invert to allow allocations larger/small than a size to
always be allowed.
* Added --alloc-fail-atstart to allow allocation failures to be disabled on
start of failgrind (then to be enabled later on with gdb or --toggle-fn)
* Added --toggle-fn that toggles the allocation failure enable/disable state
when a names function is both called and later returned from, to allow
failgrind testing to be isolated to certain parts of the program.
I think that covers most of the relevant control features from callgrind, if
you have any comments or suggestions down that road please let me know.
I'm least happy that I'm doing everything right in the instrumentation
function, so if you did want to take a quick look that would be where I'd
appreciate some feedback.
--
You are receiving this mail because:
You are watching all bug changes.
