[mips] Fix PR 21337 v1: segfault when re-reading symbols with remote debugging.

  This is a fix to a problem that was introduced with commit g3e29f34.

OK to apply?

Doug
  

On 03/31/2017 04:04 PM, Doug Gilmore wrote:
> This is a fix to a problem that was introduced with commit g3e29f34.
>
> OK to apply?
>
> Doug
>
Could a global maintainer review the patch attached to:

https://www.sourceware.org/ml/gdb-patches/2017-03/msg00559.html

when they have the chance?  Note that patch involves changing code
that is not MIPS specific.

Thanks,

Doug
  

On 03/31/2017 06:04 PM, Doug Gilmore wrote:
> This is a fix to a problem that was introduced with commit g3e29f34.
>

Which commit is that? I couldn't find it in the git tree.

Incidentally, i have a local patch that does pretty much the same thing, 
but a little further down compared to yours. It notifies gdb of objfile 
changes right before the call to read_symbols (...).

But my patch doesn't remove the original call to objfiles_changed (...). 
What is the rationale behind that change?

> diff --git a/gdb/symfile.c b/gdb/symfile.c
> index 8b79508..290b18b 100644
> --- a/gdb/symfile.c
> +++ b/gdb/symfile.c
> @@ -2592,6 +2592,12 @@ reread_symbols (void)
>  	  /* Free the obstacks for non-reusable objfiles.  */
>  	  psymbol_bcache_free (objfile->psymbol_cache);
>  	  objfile->psymbol_cache = psymbol_bcache_init ();
> +
> +	  /* Notify objfiles that we've modified objfile sections, which now
> +	     needs to be done early to ensure that, for the MIPS target,
> +	     find_pc_section won't access stale data.  PR 21337.  */

I think the PR number is not needed. After all the bug will be gone with 
this fix.
  

On 04/12/2017 11:52 AM, Luis Machado wrote:
> On 03/31/2017 06:04 PM, Doug Gilmore wrote:
>> This is a fix to a problem that was introduced with commit g3e29f34.
>>
> 
> Which commit is that? I couldn't find it in the git tree.
$ git log -n 1 3e29f34 | head
commit 3e29f34a4eef29f5b159749ccb1efb8867b2e651
Author: Maciej W. Rozycki <macro@codesourcery.com>
Date:   Fri Dec 12 13:31:53 2014 +0000

    MIPS: Keep the ISA bit in compressed code addresses
    
    1. Background information
    
    The MIPS architecture, as originally designed and implemented in
    mid-1980s has a uniform instruction word size that is 4 bytes, naturally
$ 
> 
> Incidentally, i have a local patch that does pretty much the same
> thing, but a little further down compared to yours. It notifies gdb
> of objfile changes right before the call to read_symbols (...).
That location is fine too, I just put the call at the point that
the data actually becomes stale.
> 
> But my patch doesn't remove the original call to objfiles_changed
> (...). What is the rationale behind that change?
I removed it just because would always be an duplicate call.
> 
>> diff --git a/gdb/symfile.c b/gdb/symfile.c
>> index 8b79508..290b18b 100644
>> --- a/gdb/symfile.c
>> +++ b/gdb/symfile.c
>> @@ -2592,6 +2592,12 @@ reread_symbols (void)
>>        /* Free the obstacks for non-reusable objfiles.  */
>>        psymbol_bcache_free (objfile->psymbol_cache);
>>        objfile->psymbol_cache = psymbol_bcache_init ();
>> +
>> +      /* Notify objfiles that we've modified objfile sections, which now
>> +         needs to be done early to ensure that, for the MIPS target,
>> +         find_pc_section won't access stale data.  PR 21337.  */
> 
> I think the PR number is not needed. After all the bug will be gone with this fix.
Your right, people can just run "git log -p" to recover that information.

I'll update the patch accordingly.

Thanks,

Doug