Re: [PATCH] Fix regression in -file-list-exec-source-files command.

[Doug Evans <dje@google.com>]

On Thu, Aug 19, 2010 at 3:01 AM, Joel Brobecker <brobecker@adacore.com> wrote:
>> This was tested on x86_64-linux in HEAD.  I'll see whether it's possible
>> to create a regression test or not, next.
>
> The problem that I am seeing is that, in the testsuite, we compile the
> examples using the source full filenames. For instance, using
> gdb.base/arrayidx.exp, we do:
>
>    % gcc /[...]/gdb.base/arrayidx.c -g  -lm   -o [...]
>
> As a result, the filename field ends up containing the fullname:
>
>    (gdb) interpreter-exec mi -file-list-exec-source-files
>    ^done,files=[{file="/[...]/gdb.base/arrayidx.c",fullname="/[...]/gdb.base/arrayidx.c"}, [...]
>
> Because of that, we cannot use one of our source files to demonstrate
> the problem.
>
> The only thing I can think of that we could use to demonstrate the problem
> is to check for empty filenames. Before the patch was applied, one of
> the entries looked like this:
>
>    {file="",fullname="init.c"}
>
> There is no guaranty that this is going to happen on all platforms,
> or even appear consistently on a given platform. This is probably
> dependent on the compiler as well as the way system libraries such
> as the C library have been built.  But I suppose that is better than
> nothing.

While not ideal, could you create a test in assembler/dwarf?