Re: [RFA] Ensure deterministic result order in gdb.ada/info_auto_lang.exp

[Joel Brobecker <brobecker at adacore dot com>]

> > Since both paths appear already normalized in my case, the test does fails for me and your patch
> > would fix it.  I am 99% sure I am fine with your fix, I just want to make sure we are on the same
> > page.
> A mystery to me why paths are already normalized in your build ...
> What compilation command do you see in the gdb.log for some_c.c ?

I think it depends how you configure your build, and where you do
your testing. That's why Simon asked you about it.

Personally, I build and test out-of-tree, and this is the recommended
way. And the second important element is the fact that I use an absolute
path to the configure script:

    $ cd /path/to/build
    $ /path/to/src/configure [...]
    $ make

When you configure GDB that way, then the testsuite framework will
compile the C files by passing the absolute path to those C files.
For instance on example from my gdb.log file:

    | Executing on host: gcc [...] -c -g [snip snip] /path/to/src/gdb/testsuite/gdb.base/gdb1090.c    (timeout = 300)

However, if you configure using a relative path to the configure script,
things become different. With the same example, but configured using...

    $ cd /path/to/build
    $ ../src/configure [...]

... the path given by the testsuite frame to GCC in order to compile
C files now includes these relative directory adjustments you are
seeing. Eg:

    | Executing on host: gcc [...] -c -g [...] /path/to/bld/gdb/testsuite/../../../../act/gdb-public/gdb/testsuite/gdb.base/gdb1090.c    (timeout = 300)

Whether it must be this way or not and why, I don't know. But I don't
think it's even worth exploring the idea of changing it.

-- 
Joel