[lkml]   [2004]   [Feb]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: kgdb: rename i386-stub.c to kgdb.c
Tom Rini wrote:
> On Wed, Feb 25, 2004 at 01:17:15PM -0800, George Anzinger wrote:
>>Pavel Machek wrote:
>>>>>kgdb uses really confusing names for arch-dependend parts. This fixes
>>>>>it. Okay to commit?
>>>>Why is arch/$x/kernel/$x-stub.c confusing? The name $x-stub.c is
>>>>indicative of architecture dependent code in it. Err, well so is the path.
>>>Well, looking at i386-stub.c, how do you know it is kgdb-related?
>>>>PPC and sparc stubs in present vanilla kernel use this naming convention.
>>>>That's why I adopted it.
>>>>I find kernel/kgdbstub.c, arch/$x/kernel/$x-stub.c more consistent
>>>>compared to kernel/kgdbstub.c, arch/$x/kernel/kgdb.c
>>>I actually made it kernel/kgdb.c and arch/*/kernel/kgdb.c. I believe
>>>there's no point where one could be confused....
>>gdb itself gets confused with this. Try, for example, time.c which, on the
>>x86, is in both arch and common code. I use emacs with kgdb and it gets
>>confused when I point at a location in the source and tell it to set a
>>break point.
>>Please, lets have only one of each name.
> We can't. We've had various things (most notably MODVERSIONS prior to
> 2.6) which are unhappy with that, and in the end, the problem ends up
> being fixed elsewhere. In fact, this should be able to as well. Using
> your time.c example:
> (gdb) break arch/i386/kernel/time.c:set_rtc_mmss
> Breakpoint 3 at 0xc010ee90: file arch/i386/kernel/time.c, line 174.
> (gdb) break kernel/time.c:sys_time
> Breakpoint 4 at 0xc011f0cc: file time.h, line 301.
> So shouldn't it be a matter of somehow just not having a time.c
> reference as well in the debug data?

No it is related to the point and grunt option in emacs:

I would guess it is a problem in the emacs interface where one points at a
location in the code window and enters a command to set a break point ( I think
it is "^x " (control X space)). It would appear that emacs then only sends the
file name to gdb rather than the full path.

This is not a show stopping problem, only confusing. Once gdb figures out the
right source, all is well. I usually do it by setting a break point at the
function by name, thus avoiding the point and grunt thing.

George Anzinger
Preemption patch:

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 14:01    [W:0.052 / U:43.180 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site