Re: [RFA] When getting the locno of a bpstat, handle the case of bp with null locations.

[Philippe Waroquiers <philippe.waroquiers@skynet.be>]

On Sun, 2022-11-20 at 14:34 -0500, Simon Marchi wrote:
> I ran the testsuite with the patch applied, I saw these unexpected
> failures when running with native-gdbserver or
> native-extended-gdbserver:
> 
> FAIL: gdb.base/step-over-syscall.exp: detach-on-fork=off: follow-fork=parent: break cond on target : fork: continue to marker
> FAIL: gdb.base/step-over-syscall.exp: detach-on-fork=off: follow-fork=child: break cond on target : fork: continue to marker
> 
> I bisected, and it pointed to the locno patch again.  Probably the
> expected patterns that need to be updated?
Effectively, this test contains some parts running only with gdbserver, and I missed
the needed update of the pattern.
I have posted an RFA with a fix:
[RFA] Fix step-over-syscall.exp matching regexp for $bpnum.$locno matching
https://sourceware.org/pipermail/gdb-patches/2022-November/194020.html

> 
>     continue^M
>     Continuing.^M
>     [New inferior 2 (process 1262224)]^M
>     ^M
>     Thread 1.1 "step-over-fork" hit Breakpoint 4.1, marker () at /home/simark/src/binutils-gdb/gdb/testsuite/gdb.base/step-over-fork.c:22^M
>     22      marker () {}^M
>     (gdb) FAIL: gdb.base/step-over-syscall.exp: detach-on-fork=off: follow-fork=parent: break cond on target : fork: continue to marker
> 
> However, I confirm it gets rid of the UNRESOLVEDs due to the ASan
> complaints.  But I don't have time to look at the code right now, sorry.
Thanks for testing and pointing at these regressions (and sorry about these).

There are now 3 RFA fixing some regressions
[RFA] Fix use after free introduced by $_hit_bpnum/$_hit_locno variables.
[RFA] When getting the locno of a bpstat, handle the case of bp with null locations.   
[RFA] Fix step-over-syscall.exp matching regexp for $bpnum.$locno matching

and a fix for an unrelated problem but discovered when validating the above:
[RFA] Fix jump on uninit producer_is_clang of cu.h, rm declared/undefined find_partial_die

The week-end was too busy with the above to search for the dwarf2 leaks also found while
validating the fixes. I might have time next week-end.
 
Thanks
Philippe