[RFC 0/7] [gdb/symtab, cc-with-dwz] Fix gdb.cp/breakpoint-locs.exp

[Tom de Vries]

On 9/12/23 17:23, Tom de Vries wrote:
> But that does not address the parent entry calculation.  In the simplest 
> case, the parent of an entry is the DIE parent, and it's trivial to 
> handle.  In case the DIE parent is ignored, and the parent is derived by 
> following abstract_origin chains, and we end up with a parent that is 
> not processed yet, we simply have to defer.  And if the parent is in 
> another shard, we have to defer until after the parallel for (or 
> alternatively, wait for other shards to be done, though I'm not sure 
> that level of complexity is worth it).
> 
> The problem of incorrect parent entry calculation is not very visible 
> atm.  If the parent entry calculation is incorrect, it quietly defaults 
> to no parent.  And most of the time, that doesn't result in incorrect 
> gdb behaviour.  This test-case is the exception.  Which we can work 
> around by handling PUs differently.  But we'll still quietly calculate 
> incorrect parent entries, unless we fix the generic case, as this patch 
> series proposes to do.

I managed to write a test-case that demonstrates this problem 
independently of the gdb.cp/breakpoint-locs.exp test-case, filed here ( 
https://sourceware.org/bugzilla/show_bug.cgi?id=30846 ).

Thanks,
- Tom