Re: A question about gdb script

[Michael Snyder <msnyder@vmware.com>]

Hui Zhu wrote:
> On Thu, Nov 26, 2009 at 03:15, Michael Snyder <msnyder@vmware.com> wrote:
>> Hui Zhu wrote:
>>
>>> 1. If I want get the gdb_record.xxx file of each record cycle.  It's
>>> still hard to me.  Because if I add "record save" to commands, each
>>> time it will save record entry to same file.
>>> Does gdb have some way to handle it? like:
>>> $a=1
>>> record save $a
>>> $a++
>> If you look at cmd_record_save, you see it has (char *args).
>> GDB will pass "$a" as "args", so at that point you have the
>> opportunity to say, eg. "if (args[0] == '$') then do something".
>>
>> One possibility would be to create your default name, eg.
>> gdb_record.12345, and then append $a making gdb_record.12345.1
>>
>> That's the quick and dirty way.  Something more elegant might
>> be better.
> 
> I try this way but got:
> (gdb) set $record = 3
> (gdb) record save core.$record
> Saved core file core.$record with execution log.

No, what I meant was
   (gdb) set $foo = 3
   (gdb) record save $foo

That way you can look at the *first character* of "args", and
use an 'if' to do something special with it, such as

   char *filename = strcat("gdb_record.core", args);

> Maybe I can post a patch for it. What about following:
> (gdb) set record not_overwrite_save on
> (gdb) record save core
> Saved core file core with execution log.
> (gdb) record save core
> Saved core file core.0 with execution log.
> (gdb) record save core
> Saved core file core.1 with execution log.

That's also a possibility, although I think we are headed
in the direction of having too many mode variables associated
with process rec.