[opensuse] Question on building glibc
Dear listmembers, maybe offtopic, but maybe someone could give me a pointer where to ask / what to do. I upgraded my good old SuSE 9.3 to gcc 4.1.2, I made the kernel 2.6.11 build with that compiler (was some work) and now I finally managed to make glibc-2.3.4-23.9 from the 9.3 distribution build given this combination of gcc / kernel - yes, well, with two issues: 1.) /* major issue IMHO */ one test fails, tst-signal1.c gives the output (tst-signal1.out): sending the signal now sigwait returned with Interrupted system call (4) thread joined?! child wasn't signalled The "child wasn't signalled message breaks the build, does someone has an idea what could be the root cause for this issue? Should one build in single user mode / any special precautions for glibc? 2.) /* minor issue IMHO but still not nice */ one of the math test fails too due to insufficient precision (test-float.out): testing float (without inline functions) Failure: Test: Real part of: cacosh (-2 - 3 i) == -1.9833870299165354323470769028940395 + 2.1414491111159960199416055713254211 i Result: is: -1.98338651657104492188e+00 -0x1.fbbf3800000000000000p+0 should be: -1.98338699340820312500e+00 -0x1.fbbf4000000000000000p+0 difference: 4.76837158203125000000e-07 0x1.00000000000000000000p-21 ulp : 4.0000 max.ulp : 0.0000 Failure: Test: Imaginary part of: cacosh (-2 - 3 i) == -1.9833870299165354323470769028940395 + 2.1414491111159960199416055713254211 i Result: is: 2.14144825935363769531e+00 0x1.121afa00000000000000p+1 should be: 2.14144921302795410156e+00 0x1.121b0200000000000000p+1 difference: 9.53674316406250000000e-07 0x1.00000000000000000000p-20 ulp : 4.0000 max.ulp : 0.0000 Test suite completed: 2599 test cases plus 2384 tests for exception flags executed. 2 errors occurred. I "glued" plasters over both of the failures, IMHO the second failure might be a rounding issue at 10^-6 error size - not that bad. The first error makes me headaches, but maybe someone on the list can remove it for me :-):-) Is there anything with regard to the CFLAGS that stem from rpm that should be obeyed / taken care for? _Should_ one build as super user? Thank you very much, take care Dieter Jurzitza -- ----------------------------------------------------------- | \ /\_/\ | | ~x~ |/-----\ / \ /- \_/ ^^__ _ / _ ____ / <°°__ \- \_/ | |/ | | || || _| _| _| _| if you really want to see the pictures above - use some font with constant spacing like courier! :-) ----------------------------------------------------------- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Dieter Jurzitza <dieter.jurzitza@t-online.de> writes:
Dear listmembers, maybe offtopic, but maybe someone could give me a pointer where to ask / what to do. I upgraded my good old SuSE 9.3 to gcc 4.1.2, I made the kernel 2.6.11 build with that compiler (was some work) and now I finally managed to make glibc-2.3.4-23.9 from the 9.3 distribution build given this combination of gcc / kernel - yes, well, with two issues:
1.) /* major issue IMHO */ one test fails, tst-signal1.c gives the output (tst-signal1.out): sending the signal now sigwait returned with Interrupted system call (4) thread joined?! child wasn't signalled
The "child wasn't signalled message breaks the build, does someone has an idea what could be the root cause for this issue? Should one build in single user mode / any special precautions for glibc?
This could be a kernel bug. I would really advice to update to 10.2 instead of updating those components yourself.
2.) /* minor issue IMHO but still not nice */ one of the math test fails too due to insufficient precision (test-float.out):
You can ignore those math failures. As you see those are minor details. Some ULPs are too small.
testing float (without inline functions) Failure: Test: Real part of: cacosh (-2 - 3 i) == -1.9833870299165354323470769028940395 + 2.1414491111159960199416055713254211 i Result: is: -1.98338651657104492188e+00 -0x1.fbbf3800000000000000p+0 should be: -1.98338699340820312500e+00 -0x1.fbbf4000000000000000p+0 difference: 4.76837158203125000000e-07 0x1.00000000000000000000p-21 ulp : 4.0000 max.ulp : 0.0000 Failure: Test: Imaginary part of: cacosh (-2 - 3 i) == -1.9833870299165354323470769028940395 + 2.1414491111159960199416055713254211 i Result: is: 2.14144825935363769531e+00 0x1.121afa00000000000000p+1 should be: 2.14144921302795410156e+00 0x1.121b0200000000000000p+1 difference: 9.53674316406250000000e-07 0x1.00000000000000000000p-20 ulp : 4.0000 max.ulp : 0.0000
Test suite completed: 2599 test cases plus 2384 tests for exception flags executed. 2 errors occurred.
I "glued" plasters over both of the failures, IMHO the second failure might be a rounding issue at 10^-6 error size - not that bad. The first error makes me headaches, but maybe someone on the list can remove it for me :-):-) Is there anything with regard to the CFLAGS that stem from rpm that should be obeyed / taken care for? _Should_ one build as super user?
You should not build as super user in general, Andreas -- Andreas Jaeger, Director Platform/openSUSE, aj@suse.de SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg) Maxfeldstr. 5, 90409 Nürnberg, Germany GPG fingerprint = 93A3 365E CE47 B889 DF7F FED1 389A 563C C272 A126
Dear Andreas, dear listmembers, I found the reason for the failure - without the ability to explain. For the building process I was using a special rpmrc without the "-g" option because this is needed IMHO for building the kernel based on the suse packaging concept. This missing "-g" screwed things up for glibc. If I leave "-g" in all tests (besides the floating point issue) pass smoothly, the tst-signal1 as well. If someone has an explanation, I'd be more than happy to understand :-) By the way: upgrading is not fun. New releases come with new bugs - I cannot cope with bug hunting on my main system, I needed a compiler that works better than the one I had had before. Everything beyond that like making linux-2.6.11 compile and run with gcc 4.1.2 is work, but can be decouplded from the system operation. As long as programs get shipped without having ever been run only once (10.2: xcdroast is a mess; texmacs was a mess ...) the degree of fun with upgrades is very limited. Thanks a lot for your feedback, take care Dieter Am Freitag, 27. Juli 2007 17:49 schrieb Andreas Jaeger:
Dieter Jurzitza <dieter.jurzitza@t-online.de> writes:
Dear listmembers, maybe offtopic, but maybe someone could give me a pointer where to ask / This could be a kernel bug. I would really advice to update to 10.2 instead of updating those components yourself.
-- ----------------------------------------------------------- | \ /\_/\ | | ~x~ |/-----\ / \ /- \_/ ^^__ _ / _ ____ / <°°__ \- \_/ | |/ | | || || _| _| _| _| if you really want to see the pictures above - use some font with constant spacing like courier! :-) ----------------------------------------------------------- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (2)
-
Andreas Jaeger -
Dieter Jurzitza