This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] fix PR gdb/19914


ping?

On Wed, Apr 13, 2016 at 11:18 AM, Yunlian Jiang <yunlian@google.com> wrote:
> I have a updated one. It uses separate_debug_objfile_backlink to get
> the real dwp file name if it is a debug file.
>
> --- a/gdb/dwarf2read.c
> +++ b/gdb/dwarf2read.c
> @@ -10835,13 +10835,28 @@ open_and_init_dwp_file (void)
>  {
>    struct objfile *objfile = dwarf2_per_objfile->objfile;
>    struct dwp_file *dwp_file;
> +  char *dwp_base_name;
>    char *dwp_name;
>    bfd *dbfd;
>    struct cleanup *cleanups;
>
>    /* Try to find first .dwp for the binary file before any symbolic links
>       resolving.  */
> +
> +  /* If the objfile is a debug file, find the name of the real binary
> +     file and get the name of dwp file from there. */
> +  if (objfile->separate_debug_objfile_backlink)
> +    {
> +      dwp_base_name = xstrprintf ("%s.dwp",
> +        lbasename (objfile->separate_debug_objfile_backlink->original_name));
> +    }
> +  else
> +    dwp_base_name = xstrprintf ("%s.dwp", lbasename (objfile->original_name));
> +  dwp_name = xstrprintf ("%s%s%s", ldirname(objfile->original_name),
> +                         SLASH_STRING, dwp_base_name);
> +
>    dwp_name = xstrprintf ("%s.dwp", objfile->original_name);
> +  cleanups = make_cleanup (xfree, dwp_base_name);
>    cleanups = make_cleanup (xfree, dwp_name);
>
>    dbfd = open_dwp_file (dwp_name);
>
> On Fri, Apr 8, 2016 at 3:54 PM, Doug Evans <dje@google.com> wrote:
>> Yunlian Jiang writes:
>>  > Hi.
>>  >
>>  > This tries to fix PR gdb/19914. When the object file is not
>>  > a binary but a .debug file, the function to try to find the .dwp
>>  > file in the debug directory does not work. This one tries to solve
>>  > this problem.
>>  >
>>  > 2016-04-05  Yunlian Jiang <yunlian@google.com>
>>  >                * gdb/dwarf2read.c: Try to find .dwp for .debug
>>  >
>>  >
>>  > diff --git a/gdb/dwarf2read.c b/gdb/dwarf2read.c
>>  > index dcd49e3..551376a 100644
>>  > --- a/gdb/dwarf2read.c
>>  > +++ b/gdb/dwarf2read.c
>>  > @@ -10854,6 +10854,18 @@ open_and_init_dwp_file (void)
>>  >        dbfd = open_dwp_file (dwp_name);
>>  >      }
>>  >
>>  > +  if (dbfd == NULL && strlen(dwp_name) > 10)
>>  > +    {
>>  > +      /* Try to find .dwp for the debug file.  */
>>  > +      if (!strcmp(dwp_name + strlen(dwp_name) - 10, ".debug.dwp"))
>>  > +        {
>>  > +           dwp_name[strlen(dwp_name) - 10] = '\0';
>>  > +           dwp_name = xstrprintf ("%s.dwp", dwp_name);
>>  > +           make_cleanup (xfree, dwp_name);
>>  > +           dbfd = open_dwp_file (dwp_name);
>>  > +        }
>>  > +    }
>>  > +
>>  >    if (dbfd == NULL)
>>  >      {
>>  >        if (dwarf_read_debug)
>>
>> Hi.
>>
>> I've looked at this a little bit.
>> I don't have time for a complete review at the moment,
>> but one thing I'm wondering about is whether this extra
>> test should go in open_dwp_file instead.
>>
>> Plus, while we're making the assumption that the separate debug
>> file ends with ".debug" that's not necessarily true.
>> Not sure we can do better, but one thought is to look in
>> separate_debug_objfile_backlink and use that.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]