[PATCH] set saner MAX_ARGS

[Dmitry V. Levin]

On Fri, Aug 19, 2011 at 05:38:19PM +0200, Denys Vlasenko wrote:
> On Fri, 2011-08-19 at 00:02 +0400, Dmitry V. Levin wrote:
> > On Thu, Aug 18, 2011 at 12:47:38PM +0200, Denys Vlasenko wrote:
[...]
> > > > Alternative solution is to make sys_sigreturn print mask
> > > > on entry, not on exit. What is the reson it doesn't do that now?
> > 
> > Can sys_sigreturn be interrupted somehow?
> 
> Does it matter?
> 
> Currently, sigreturn in strace looks like this:
> 
> wait4(-1, 0xbfa81764, WSTOPPED, NULL)   = ? ERESTARTSYS (To be restarted)
> --- {si_signo=SIGINT, si_code=SI_USER, si_pid=13194, si_uid=0} (Interrupt) ---
> sigreturn()                             = ? (mask now [])
> --- {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=13194, si_status=0, si_utime=0, si_stime=0} (Child exited) ---
> sigreturn()                             = ? (mask now [])
> wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], WSTOPPED, NULL) = 13194
> 
> 
> I propose to change it into something like this:
> 
> wait4(-1, 0xbfa81764, WSTOPPED, NULL)   = ? ERESTARTSYS (To be restarted)
> --- {si_signo=SIGINT, si_code=SI_USER, si_pid=13194, si_uid=0} (Interrupt) ---
> sigreturn([])                           = 0
> --- {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=13194, si_status=0, si_utime=0, si_stime=0} (Child exited) ---
> sigreturn([])                           = 0
> wait4(-1, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], WSTOPPED, NULL) = 13194
> 
> because that's how it _really_ happens: the signal mask is actually known
> BEFORE sigreturn is executed, so why are we printing it AFTER it returns?
> (And in the process, torture ourselves with stupid saving/restoring of
> sigmask for that). Look at sys_sigreturn() in strace source to see it.

If the signal mask is known on enter, lets print it also on enter.

> (The only wart here is that sigmask is not a sigreturn's parameter, so
> printing it as "sigreturn([])" is slightly wrong/misleading. Any ideas
> how to show it better? With curly braces maybe - {[mask]}?)
> 
> Do you like this idea?

Curly braces usually mean a pointer to a structure, but I see no better
way to express the signal mask on enter.


-- 
ldv
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20110819/b6c49b78/attachment.bin>