Re: [RFC] breakpoints/19474 [was Re: RFC: branching for GDB 7.11 soon? (possibly Wed)]

[Yao Qi <qiyaoltc at gmail dot com>]

[I am back from vacation...]

Hi Keith,
Thanks for looking at this bug.

On 04/02/16 00:45, Keith Seitz wrote:
> It is also here where we know we are attempting to fetch a list of
> symtabs for the default source file, meaning all symtabs with the "same"
> file name as default symtab.
>
> So the simple (but apparently working?) algorithm simply compares the
> default symtab's fullname vs a reconstruction of the SYMTAB_DIRNAME and
> basename of the "collected" symtab. [It also considers substitute-path.]

The algorithm/rationale sounds good to me.  I had a similar fix in my
side, but it causes regressions in the following test cases, in which
DW_AT_comp_dir is empty or incorrect.

 gdb.arch/amd64-entry-value-param.exp
 gdb.cp/namelessclass.exp
 gdb.dwarf2/dw2-common-block.exp
 gdb.dwarf2/dw2-single-line-discriminators.exp

but your patch looks better.

> I would consider this fairly risky for inclusion into a release without
> sufficient testing in HEAD, but when it comes to releases, I tend to be
> quite conservative.

Yes, I agree.  The problem itself can't be fixed properly without a big
structural change.  We don't have to ship this patch to 7.11, IMO.  It
is OK for mainline, however.

> I've appended the patch below.  This patch causes no regressions in the
> test suite, and it fixes 19474.

> I'm sure there are corner cases and a whole bunch of other problems with
> this approach, but at least it is isolated to one place (for better or
> worse).
>
> Anyone have a better idea?

I don't have any.  I agree that the directory name comparison in you
next mail is better.

--
Yao (éå)