Hi Gert, Hi Steven, Gert Wollny, on 2020-08-12 12:25:31 +0200: > Am Dienstag, den 11.08.2020, 08:40 -0500 schrieb Steven Robbins: > > On my i386 chroot build, there are even more failures: > [...] > > So far these failures were usually coming from using the old FPU on > i386, which results in in 80 bit intermediate accuracy (that' what the > -fstore-float flag tries to quell), and then the expected test results > were off a bit. Maybe we should just enable sse or even sse2, I doubt > that anyone is running this library on i386 hardware that is not sse2 > capable. I erroneously interrupted the build on my Atom CPU (actually it took me a while to figure out that I should specify option --no-parallel to dh avoid the drive thrashing like hell and ultimately OOM conditions mid-way during the build, hence the delay). I'm about to restart a run: - without patches mentioned earlier, - with -msse replacing -fstore-float in i386 specific build options, - with ccache, so I gain some checkpoints if I screw the build again. Hope it will reveal something interesting, but count some days to it. Maybe a virtual machine would be faster in this case; 1 GiB of RAM is considered quite light nowadays. Kind Regards, -- Étienne Mollier <etienne.mollier@mailoo.org> Old rsa/3072: 5ab1 4edf 63bb ccff 8b54 2fa9 59da 56fe fff3 882d New rsa/4096: 8f91 b227 c7d6 f2b1 948c 8236 793c f67e 8f0d 11da Sent from /dev/tty1, please excuse my verbosity.
Attachment:
signature.asc
Description: PGP signature