[PATCH 3/4] decode_select: Decode timeout for interrupted syscall

[Dmitry V. Levin]

On Mon, Jul 28, 2008 at 03:42:33PM -0700, Roland McGrath wrote:
> > 	* desc.c (decode_select): Decode timeout for interrupted syscall.
> 
> The user's timeval pointer only gets updated in Linux's special
> incompatible mode ("personality" bits without STICKY_TIMEOUTS), which
> noone ever uses in reality.  The timeout value there is the original one
> already printed at syscall entry regardless of how long select did or
> didn't block, so it's just confusingly inaccurate to print it again
> labelled "left".

I wouldn't say that this mode is no longer used in reality.
Nevertheless, since this change may lead to confusing output,
I have to recall it.

> > 	* syscall.c (trace_syscall): Fix ERESTART* decode output.
> 
> Why is it better?  In those cases, noone ever sees a -1.
> Either it later becomes EINTR, or it doesn't really return.

I agree.


-- 
ldv
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://lists.strace.io/pipermail/strace-devel/attachments/20080903/2006e809/attachment.bin>