On Fri, 2013-03-15 at 15:36 +0100, Jan Kratochvil wrote:
On Fri, 15 Mar 2013 14:37:53 +0100, Mark Wielaard wrote:
The only reason I was hesitant about the patch was the testcase. I assume this mimics what dwfl_linux_proc_report () would do, and it was easier to write it using dwfl_report_elf ()?
dwfl_linux_proc_report uses live PID but if you have a dead core file how to pass a PID to dwfl_linux_proc_report?
It is also explained in the patch mail that elfutils uses dwfl_report_segment which does not have such bug but it is elfutils internal function. Applications cannot use it. Maybe the unwinder could have a different API without using dwfl_report_elf but even in such case dwfl_report_elf exists so its bug should be fixed (by this patch).
Yeah, it seems good to fix this. I was just wondering if there was some other way to trigger it. BTW dwfl_report_segment is a public function.
If you could add a comment where the constants for base (0x7f3560c92000) and funcaddr (0x7f3560c92585) come from (and/or how to calculate your own in case someone wants to extend the testcase) that would be appreciated.
Done.
Therefore I expect it can be checked in.
Yes, thanks for adding the explanation to the testcase.
BTW. For the git commit message it is nice if the first line (< 78 chars) is a short summary of the fix. That way a git shortlog looks a little easier to read.
Cheers,
Mark