[RFC,9] Revert basenames_may_differ patch

  Commit a0c1ffedc regressed certain cases coming from Eclipse.
See PR breakpoints/24915.

This patch reverts the commit for the gdb 9 release.

gdb/ChangeLog
2020-02-05  Tom Tromey  <tromey@adacore.com>

	PR breakpoints/24915:
	* source.c (find_and_open_source): Do not check basenames_may_differ.

gdb/testsuite/ChangeLog
2020-02-05  Tom Tromey  <tromey@adacore.com>

	PR breakpoints/24915:
	* gdb.base/annotate-symlink.exp: Use setup_xfail.

Change-Id: Iadbf42f35eb40c95ad32b2108ae25d8f199998bd
---
 gdb/ChangeLog                               |  5 +++++
 gdb/source.c                                | 19 +++++++------------
 gdb/testsuite/ChangeLog                     |  5 +++++
 gdb/testsuite/gdb.base/annotate-symlink.exp |  3 +++
 4 files changed, 20 insertions(+), 12 deletions(-)
  

Hi Tom,

On Wed, Feb 05, 2020 at 10:55:15AM +0100, Tom Tromey wrote:
> Commit a0c1ffedc regressed certain cases coming from Eclipse.
> See PR breakpoints/24915.
> 
> This patch reverts the commit for the gdb 9 release.
> 
> gdb/ChangeLog
> 2020-02-05  Tom Tromey  <tromey@adacore.com>
> 
> 	PR breakpoints/24915:
> 	* source.c (find_and_open_source): Do not check basenames_may_differ.
> 
> gdb/testsuite/ChangeLog
> 2020-02-05  Tom Tromey  <tromey@adacore.com>
> 
> 	PR breakpoints/24915:
> 	* gdb.base/annotate-symlink.exp: Use setup_xfail.

Thanks for the patch. I think this is the most reasonable thing
to do. Can you apply it?

> 
> Change-Id: Iadbf42f35eb40c95ad32b2108ae25d8f199998bd
> ---
>  gdb/ChangeLog                               |  5 +++++
>  gdb/source.c                                | 19 +++++++------------
>  gdb/testsuite/ChangeLog                     |  5 +++++
>  gdb/testsuite/gdb.base/annotate-symlink.exp |  3 +++
>  4 files changed, 20 insertions(+), 12 deletions(-)
> 
> diff --git a/gdb/source.c b/gdb/source.c
> index 2497e7f750e..a737c95e7a4 100644
> --- a/gdb/source.c
> +++ b/gdb/source.c
> @@ -1025,10 +1025,7 @@ find_and_open_source (const char *filename,
>        result = gdb_open_cloexec (fullname->get (), OPEN_MODE, 0);
>        if (result >= 0)
>  	{
> -	  if (basenames_may_differ)
> -	    *fullname = gdb_realpath (fullname->get ());
> -	  else
> -	    *fullname = gdb_abspath (fullname->get ());
> +	  *fullname = gdb_realpath (fullname->get ());
>  	  return scoped_fd (result);
>  	}
>  
> @@ -1072,12 +1069,9 @@ find_and_open_source (const char *filename,
>    if (rewritten_filename != NULL)
>      filename = rewritten_filename.get ();
>  
> -  openp_flags flags = OPF_SEARCH_IN_PATH;
> -  if (basenames_may_differ)
> -    flags |= OPF_RETURN_REALPATH;
> -
>    /* Try to locate file using filename.  */
> -  result = openp (path, flags, filename, OPEN_MODE, fullname);
> +  result = openp (path, OPF_SEARCH_IN_PATH | OPF_RETURN_REALPATH, filename,
> +		  OPEN_MODE, fullname);
>    if (result < 0 && dirname != NULL)
>      {
>        /* Remove characters from the start of PATH that we don't need when
> @@ -1098,15 +1092,16 @@ find_and_open_source (const char *filename,
>        cdir_filename.append (SLASH_STRING);
>        cdir_filename.append (filename_start);
>  
> -      result = openp (path, flags, cdir_filename.c_str (), OPEN_MODE,
> -		      fullname);
> +      result = openp (path, OPF_SEARCH_IN_PATH | OPF_RETURN_REALPATH,
> +		      cdir_filename.c_str (), OPEN_MODE, fullname);
>      }
>    if (result < 0)
>      {
>        /* Didn't work.  Try using just the basename.  */
>        p = lbasename (filename);
>        if (p != filename)
> -	result = openp (path, flags, p, OPEN_MODE, fullname);
> +	result = openp (path, OPF_SEARCH_IN_PATH | OPF_RETURN_REALPATH, p,
> +			OPEN_MODE, fullname);
>      }
>  
>    return scoped_fd (result);
> diff --git a/gdb/testsuite/gdb.base/annotate-symlink.exp b/gdb/testsuite/gdb.base/annotate-symlink.exp
> index d22593ca364..830949daa42 100644
> --- a/gdb/testsuite/gdb.base/annotate-symlink.exp
> +++ b/gdb/testsuite/gdb.base/annotate-symlink.exp
> @@ -47,5 +47,8 @@ gdb_breakpoint func message
>  
>  gdb_test_no_output "set annotate 1"
>  
> +# The patch to cause this output was reverted.
> +# See PR breakpoints/24915.
> +setup_xfail *-*-* 24915
>  gdb_test "continue" \
>      "Breakpoint .* func .*realname-expand-link.c:$decimal\r\n\032\032.*realname-expand-link.c:.*"
> -- 
> 2.21.1
  

>>>>> "Joel" == Joel Brobecker <brobecker@adacore.com> writes:

Joel> Thanks for the patch. I think this is the most reasonable thing
Joel> to do. Can you apply it?

I'm going to apply it to the gdb 9 branch shortly.  I suppose I will
apply to master as well and then open a bug about the problem.

Tom
  

On 2020-02-07 2:38 p.m., Tom Tromey wrote:
>>>>>> "Joel" == Joel Brobecker <brobecker@adacore.com> writes:
> 
> Joel> Thanks for the patch. I think this is the most reasonable thing
> Joel> to do. Can you apply it?
> 
> I'm going to apply it to the gdb 9 branch shortly.  I suppose I will
> apply to master as well and then open a bug about the problem.
> 
> Tom

I had a prototype of a fix here:

  https://github.com/simark/binutils-gdb/commits/symtab-realpath

My idea is simply that we should keep both versions of the symtab path:

- an absolute path where we have not resovled the symlinks and all (i.e. a
  non-canonical version)
- a path passed through realpath, so where the symlinks and ".." have been resolved
  (i.e a canonical version)

Different parts of GDB would use the one they prefer.  It is described in more
details here:

https://sourceware.org/ml/gdb-patches/2020-02/msg00008.html

If you think this is a good direction, I could clean up the patch and submit
it for real (after some more testing to make sure it actually does what we
want).

I don't remember exactly, but I think that my patch ends up removing your
change anyway, because it goes back to always computing the realpath, regardless
of basenames_may_differ (on top of also keeping the non-real absolute path around).

Simon