Re: Fix gdb.ada/bp_c_mixed_case.exp (PR gdb/22670) (Re: [PATCH 3/3] Add new gdb.ada/bp_c_mixed_case testcase for PR gdb/22670)

[Pedro Alves <palves at redhat dot com>]

On 01/08/2018 03:57 AM, Joel Brobecker wrote:
> On Fri, Jan 05, 2018 at 04:34:39PM +0000, Pedro Alves wrote:

>> Below's a fix for this one.
> 
> Thanks!
> 
> I confirm the test now passes for me as well :). 

Great!

> I have a question though:
> 
>> >From 439f8c51ff8f6cd9fb3bbc330a40492a15992add Mon Sep 17 00:00:00 2001
>> From: Pedro Alves <palves@redhat.com>
>> Date: Fri, 5 Jan 2018 00:17:19 +0000
>> Subject: [PATCH 1/2] Fix gdb.ada/bp_c_mixed_case.exp (PR gdb/22670)
>>
>> The problem here is that we were using the user-provided lookup name
>> literally for linkage name comparisons.  I.e., "<MixedCase>" with the
>> "<>"s included.  That obviously can't work since the "<>" are not
>> really part of the linkage name.  The original idea was that we'd use
>> the symbol's language to select the right symbol name matching
>> algorithm, but that doesn't work for Ada because it's not really
>> possible to unambiguously tell from the linkage name alone whether
>> we're dealing with Ada symbols, so Ada minsyms end up with no language
>> set, or sometimes C++ set.  So fix this by treating Ada mode specially
>> when determining the linkage name to match against.
> 
> I am wondering why minimal symbols are involved in this case,
> considering that the C file was build with debugging information.
> Shouldn't we be getting the function's address from the partial/full
> symtabs instead?

AFAIK, GDB always worked this way for linespecs, even before my C++
wildmatching patches -- we collect symbols from both debug info and
minsyms, and coalesce them by address to avoid duplicates
(linespec.c:add_matching_symbols_to_info).  

The completion paths then try to do the same thing (though implemented
differently) [1].

I think it makes sense because:

- even if you have debug information in the binary, the debug information
  won't cover all function symbols.  Some may be internal linker-/compiler-
  generated symbols.  Or..

- ..there may be multiple symbols with the same name in different
  compilation units that all end up in the same binary/objfile.  Some may
  have debug info while others not.  E.g. (C, but applies to any language,
  or mixed languages):

static void function () {}                // in file1.c, compiled with -g
static int function (int p) { return p; } // in file2.c, compiled WITHOUT -g

  I could easily see the 'with'/'without -g' functions ending up both in
  the same objfile via static linking, for example.  We want "b <function>" to
  set a breakpoint on all the functions.

[1] - the C++ wildmatching series eliminated the divergence a bit, but
there's still a lot more duplication that there should ideally be.
One of the points of the gdb.linespec/cpcompletion.exp and
gdb.linespec/cpls-*.exp testcases is making sure that completion
and actually setting a breakpoint finds the same locations.

Let me know if this answers your question.

Thanks,
Pedro Alves