Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sat, 12 Jan 2013 14:55:34 +0100
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: SIGALRM and new 'done' method

I could reproduce the initial SIGALRM problem we had before 3040fc9 by forcing OS_TIMER=0 in arch.h but I can't reproduce this one. And I really can't see how 4a7c6df could fail to resolve the issue.

Does this only ever happen when testing several (ie. all) formats? If so, that is likely a clue.

magnum


On 12 Jan, 2013, at 1:46 , Claudio André <claudioandre.br@...il.com> wrote:

> I had problems using both. Even the third option below fails (seems it can works sometimes, but not always).
> 
> 
> - common-opencl.c
> void opencl_process_event(void)
> {
>    static int bench_cludge = 0;
> 
>    /* bench_running may be reset while we still have the benchmark
>       timer active, leading to SIGALRM. Only seen with !OS_TIMER. */
>    bench_cludge |= bench_running;
> 
>    if (!bench_cludge) {
> #if !OS_TIMER
>        sig_timer_emu_tick();
> #endif
>        if (event_pending) {
> ...
> 
> - bench.c
> static void bench_handle_timer(int signum)
> {
>    bench_running = 0;
> #ifndef SA_RESTART
>    signal(SIGALRM, bench_handle_timer);
> #endif
> }
> 
> Em 11-01-2013 22:16, magnum escreveu:
>> On 12 Jan, 2013, at 0:57 , Claudio André<claudioandre.br@...il.com>  wrote:
>>> Em 11-01-2013 21:38, magnum escreveu:
>>>> Try reverting your hack and apply f8bf37d instead. Still a hack, but a better one (I hope).
>>>> 
>>>> magnum
>>> Seems to be ok.
>> Please try this one too. If it works, I think it as a better fix.
>> 
>> magnum
>> 
> 
> 


Powered by blists - more mailing lists

Your e-mail address:

Powered by Openwall GNU/*/Linux - Powered by OpenVZ