xcoffread.c: Fix -Werror=dangling-pointer= issue with main_subfile.

  GCC 13 points out that main_subfile has local function scope, but a
pointer to it is assigned to the global inclTable array subfile
element field:

In function ‘void process_linenos(CORE_ADDR, CORE_ADDR)’,
    inlined from ‘void aix_process_linenos(objfile*)’ at xcoffread.c:727:19,
    inlined from ‘void aix_process_linenos(objfile*)’ at xcoffread.c:720:1:
xcoffread.c:629:37: error: storing the address of local variable ‘main_subfile’ in ‘*inclTable.19_45 + _28._inclTable::subfile’ [-Werror=dangling-pointer=]
  629 |               inclTable[ii].subfile = &main_subfile;
      |               ~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~
xcoffread.c: In function ‘void aix_process_linenos(objfile*)’:
xcoffread.c:579:18: note: ‘main_subfile’ declared here
  579 |   struct subfile main_subfile;
      |                  ^~~~~~~~~~~~
xcoffread.c:496:19: note: ‘inclTable’ declared here
  496 | static InclTable *inclTable;    /* global include table */
      |                   ^~~~~~~~~

Fix this by making main_subfile file static that is allocated and
deallocated together with inclTable and allocate_include_entry and
xcoff_symfile_finish. Adjust the use of main_subfile in
process_linenos to take a pointer to the struct subfile.
---
 gdb/xcoffread.c | 24 +++++++++++++-----------
 1 file changed, 13 insertions(+), 11 deletions(-)
  

On 4/29/23 17:13, Mark Wielaard wrote:
> GCC 13 points out that main_subfile has local function scope, but a
> pointer to it is assigned to the global inclTable array subfile
> element field:
> 
> In function ‘void process_linenos(CORE_ADDR, CORE_ADDR)’,
>     inlined from ‘void aix_process_linenos(objfile*)’ at xcoffread.c:727:19,
>     inlined from ‘void aix_process_linenos(objfile*)’ at xcoffread.c:720:1:
> xcoffread.c:629:37: error: storing the address of local variable ‘main_subfile’ in ‘*inclTable.19_45 + _28._inclTable::subfile’ [-Werror=dangling-pointer=]
>   629 |               inclTable[ii].subfile = &main_subfile;
>       |               ~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~
> xcoffread.c: In function ‘void aix_process_linenos(objfile*)’:
> xcoffread.c:579:18: note: ‘main_subfile’ declared here
>   579 |   struct subfile main_subfile;
>       |                  ^~~~~~~~~~~~
> xcoffread.c:496:19: note: ‘inclTable’ declared here
>   496 | static InclTable *inclTable;    /* global include table */
>       |                   ^~~~~~~~~
> 
> Fix this by making main_subfile file static that is allocated and
> deallocated together with inclTable and allocate_include_entry and
> xcoff_symfile_finish. Adjust the use of main_subfile in
> process_linenos to take a pointer to the struct subfile.

I'm not familiar at all with this code, but your change looks reasonable
to me.

Some style comments:

> ---
>  gdb/xcoffread.c | 24 +++++++++++++-----------
>  1 file changed, 13 insertions(+), 11 deletions(-)
> 
> diff --git a/gdb/xcoffread.c b/gdb/xcoffread.c
> index d71127b40f6..db6f2df6c0a 100644
> --- a/gdb/xcoffread.c
> +++ b/gdb/xcoffread.c
> @@ -498,6 +498,9 @@ static int inclIndx;		/* last entry to table */
>  static int inclLength;		/* table length */
>  static int inclDepth;		/* nested include depth */
>  
> +/* subfile structure for the main compilation unit.  */
> +static struct subfile *main_subfile;

Remove "struct".

> @@ -548,6 +551,7 @@ allocate_include_entry (void)
>        inclTable = XCNEWVEC (InclTable, INITIAL_INCLUDE_TABLE_LENGTH);
>        inclLength = INITIAL_INCLUDE_TABLE_LENGTH;
>        inclIndx = 0;
> +      main_subfile = new (struct subfile);

"new subfile" would be enough.

>    for (int ii = 0; ii < inclIndx; ++ii)
>      {
> -      if (inclTable[ii].subfile != ((struct subfile *) &main_subfile)
> +      if (inclTable[ii].subfile != ((struct subfile *) main_subfile)

Since you touch this line, I think you could remove the cast.
main_subfile is already of the right type.

Simon
  

Hi Simon,

On Mon, May 01, 2023 at 09:53:39AM -0400, Simon Marchi wrote:
> > Fix this by making main_subfile file static that is allocated and
> > deallocated together with inclTable and allocate_include_entry and
> > xcoff_symfile_finish. Adjust the use of main_subfile in
> > process_linenos to take a pointer to the struct subfile.
> 
> I'm not familiar at all with this code, but your change looks reasonable
> to me.

I am also not familiar with this code, but noticed that the
fedora-latest and suse-tumbleweed buildbots started failing when they
upgraded to GCC 13.1. The simplest solution seemed to be to give both
structures the same lifetime. I didn't see any regressions with this
fix (but don't actually know if the testsuite triggers this code).

> Some style comments:
> 
> > ---
> >  gdb/xcoffread.c | 24 +++++++++++++-----------
> >  1 file changed, 13 insertions(+), 11 deletions(-)
> > 
> > diff --git a/gdb/xcoffread.c b/gdb/xcoffread.c
> > index d71127b40f6..db6f2df6c0a 100644
> > --- a/gdb/xcoffread.c
> > +++ b/gdb/xcoffread.c
> > @@ -498,6 +498,9 @@ static int inclIndx;		/* last entry to table */
> >  static int inclLength;		/* table length */
> >  static int inclDepth;		/* nested include depth */
> >  
> > +/* subfile structure for the main compilation unit.  */
> > +static struct subfile *main_subfile;
> 
> Remove "struct".

OK.
 
> > @@ -548,6 +551,7 @@ allocate_include_entry (void)
> >        inclTable = XCNEWVEC (InclTable, INITIAL_INCLUDE_TABLE_LENGTH);
> >        inclLength = INITIAL_INCLUDE_TABLE_LENGTH;
> >        inclIndx = 0;
> > +      main_subfile = new (struct subfile);
> 
> "new subfile" would be enough.

OK.

> >    for (int ii = 0; ii < inclIndx; ++ii)
> >      {
> > -      if (inclTable[ii].subfile != ((struct subfile *) &main_subfile)
> > +      if (inclTable[ii].subfile != ((struct subfile *) main_subfile)
> 
> Since you touch this line, I think you could remove the cast.
> main_subfile is already of the right type.

Right.

Will sent v2 with those changes.

Thanks,

Mark