Bug#642112: [PATCH] timers for kFreeBSD

2012-04-01 Thread Robert Millan
El 1 d’abril de 2012 0:02, Robert Millan r...@debian.org ha escrit: I will test older kernels and port it to amd64. I added backward compatibility for kfreebsd-8.x. Tested with 8.1, 8.3~ and 9.0. In all cases it passes the full tst-timer* testsuite. This is an improvement over tst-timer4 and

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-31 Thread Robert Millan
El 31 de març de 2012 1:16, Robert Millan r...@debian.org ha escrit: Somewhat.  My latest patch (after fixing other problems [1]) now passes tst-timer [2], but only *sometimes* (looks like a race).  I isolated my code into a standalone app by renaming the timer_* functions and found that when

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-30 Thread Robert Millan
El 30 de març de 2012 9:11, Petr Salinger petr.salin...@seznam.cz ha escrit: /* Signal needed for the kernel-supported POSIX timer implementation.  We can reuse the cancellation signal since we can distinguish  cancellation from timer expirations.  */ #define SIGTIMER        SIGCANCEL Are

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-26 Thread Robert Millan
El 26 de març de 2012 0:14, Robert Millan r...@debian.org ha escrit: 9.0/i386 kernel running i386 binary - GOOD 10.0/i386 kernel running i386 binary - GOOD Sorry but my i386 tests were wrong, I did them while using unpatched libc (sigh). Here are some proper tests for i386 (same result using

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-25 Thread Christoph Egger
Hi! Robert Millan r...@debian.org writes: Attached patch adds a new implementation of timers for kFreeBSD. It is heavily based on the Linux/NPTL counterpart. I've verified that this fixes #642112 (timer_settime fails on kfreebsd when timeout approaches max(time_t)) and #662018 (No support

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-25 Thread Christoph Egger
Hi! Robert Millan r...@debian.org writes: El 25 de març de 2012 21:28, Christoph Egger christ...@debian.org ha escrit: It fails some regression tests on eglibc, is that on purpose? For tst-timer* I got the same failures with sid chroot on kfreebsd 8.1, however they work fine with

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-25 Thread Robert Millan
El 25 de març de 2012 21:28, Christoph Egger christ...@debian.org ha escrit: It fails some regression tests on eglibc, is that on purpose? For tst-timer* I got the same failures with sid chroot on kfreebsd 8.1, however they work fine with kfreebsd-10. I don't have kfreebsd 8.3 and 9.0 handy atm

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-25 Thread Robert Millan
El 25 de març de 2012 23:20, Christoph Egger christ...@debian.org ha escrit: My run was on latest kfreebsd-9 kernel from sid. I can test on my 8.3 systems later, though we'll need to get this working on 8.1 as that's what the buildds are running. Sorry I got confused. It works on all versions

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-25 Thread Robert Millan
El 26 de març de 2012 0:14, Robert Millan r...@debian.org ha escrit: Sorry I got confused.  It works on all versions when it is built with i386 libc. I think I know why, but can't fix it right now. timer_t is abused to store a pointer, but its size is only int32_t. timer_create() sets it

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-25 Thread Christoph Egger
Hi! Robert Millan r...@debian.org writes: El 25 de març de 2012 23:20, Christoph Egger christ...@debian.org ha escrit: My run was on latest kfreebsd-9 kernel from sid. I can test on my 8.3 systems later, though we'll need to get this working on 8.1 as that's what the buildds are running.

Bug#642112: [PATCH] timers for kFreeBSD

2012-03-24 Thread Robert Millan
Hi, Attached patch adds a new implementation of timers for kFreeBSD. It is heavily based on the Linux/NPTL counterpart. I've verified that this fixes #642112 (timer_settime fails on kfreebsd when timeout approaches max(time_t)) and #662018 (No support for CLOCK_MONOTONIC). It also addresses a