Proper stack alignment for constructors on i386

2012-12-22 Thread Mark Kettenis
So the unaligned thread stack issue on i386 made me look at stack alignment in constructors, and I noticed that things weren't quite right there either. Rather than doing arithmetic and depending on the stack being aligned properly, just force the alignment to be right just before we start executi

Re: evince segfaults in sse2_composite_over_n_8_8888, i386, Dec 12 snapshot

2012-12-22 Thread Florian Obser
On Sat, Dec 22, 2012 at 08:33:47PM +0100, Mark Kettenis wrote: > > Date: Sat, 22 Dec 2012 20:19:17 +0100 (CET) > > From: Mark Kettenis > > > > > Date: Sat, 22 Dec 2012 18:36:53 + > > > From: Florian Obser > > > > > > 0x0298c256 : movdqa > > > %xmm0,0xffc8(%ebp) > > > > I expected an

Re: evince segfaults in sse2_composite_over_n_8_8888, i386, Dec 12 snapshot

2012-12-22 Thread Matthieu Herrb
On Sat, Dec 22, 2012 at 08:33:47PM +0100, Mark Kettenis wrote: > > Date: Sat, 22 Dec 2012 20:19:17 +0100 (CET) > > From: Mark Kettenis > > > > > Date: Sat, 22 Dec 2012 18:36:53 + > > > From: Florian Obser > > > > > > 0x0298c256 : movdqa > > > %xmm0,0xffc8(%ebp) > > > > I expected an

RFC: add nl(1)

2012-12-22 Thread Arto Jonsson
Hi, this patch adds nl(1) from netbsd which, if I've understood correctly, is required by posix 2008. I've made the following changes to the code from netbsd: * use strtonum * getprogname() -> __progname * remove old RCID * use __dead * clean up includes For the manual: * remove .Pp before .Bl

Re: evince segfaults in sse2_composite_over_n_8_8888, i386, Dec 12 snapshot

2012-12-22 Thread Mark Kettenis
> Date: Sat, 22 Dec 2012 20:19:17 +0100 (CET) > From: Mark Kettenis > > > Date: Sat, 22 Dec 2012 18:36:53 + > > From: Florian Obser > > > > 0x0298c256 : movdqa > > %xmm0,0xffc8(%ebp) > > I expected an instruction like that. And since we have > > > ebp0x82049574 0x

Re: evince segfaults in sse2_composite_over_n_8_8888, i386, Dec 12 snapshot

2012-12-22 Thread Mark Kettenis
> Date: Sat, 22 Dec 2012 18:36:53 + > From: Florian Obser > > 0x0298c256 : movdqa %xmm0,0xffc8(%ebp) I expected an instruction like that. And since we have > ebp0x82049574 0x82049574 the address 0xffc8(%ebp) isn't 16-byte aligned. So it looks like our thread st

Re: evince segfaults in sse2_composite_over_n_8_8888, i386, Dec 12 snapshot

2012-12-22 Thread Matthieu Herrb
On Sat, Dec 22, 2012 at 07:08:45PM +0100, Mark Kettenis wrote: > > Date: Sat, 22 Dec 2012 18:45:31 +0100 > > From: Matthieu Herrb > > List-Owner: > > > > On Fri, Dec 21, 2012 at 05:45:11PM +0100, Giovanni Bechis wrote: > > > > [moving to tech@ since it's not a por

Re: evince segfaults in sse2_composite_over_n_8_8888, i386, Dec 12 snapshot

2012-12-22 Thread Florian Obser
On Sat, Dec 22, 2012 at 07:08:45PM +0100, Mark Kettenis wrote: > > Date: Sat, 22 Dec 2012 18:45:31 +0100 > > From: Matthieu Herrb > > List-Owner: > > > > On Fri, Dec 21, 2012 at 05:45:11PM +0100, Giovanni Bechis wrote: > > > > [moving to tech@ since it's not a por

Re: evince segfaults in sse2_composite_over_n_8_8888, i386, Dec 12 snapshot

2012-12-22 Thread Mark Kettenis
> Date: Sat, 22 Dec 2012 18:45:31 +0100 > From: Matthieu Herrb > List-Owner: > > On Fri, Dec 21, 2012 at 05:45:11PM +0100, Giovanni Bechis wrote: > > [moving to tech@ since it's not a ports issue] > > > On 12/21/12 17:40, Florian Obser wrote: > > > Hi, > > > evin

Re: evince segfaults in sse2_composite_over_n_8_8888, i386, Dec 12 snapshot

2012-12-22 Thread Matthieu Herrb
On Fri, Dec 21, 2012 at 05:45:11PM +0100, Giovanni Bechis wrote: [moving to tech@ since it's not a ports issue] > On 12/21/12 17:40, Florian Obser wrote: > > Hi, > > evince-3.6.1 segfaults with this pdf: > > http://gowers.files.wordpress.com/2012/02/elsevierstatementfinal.pdf > > > > I tried a k