Running Hercules emerge(d) 3.01 with CFLAGS="-O2 -march=athlon-xp -pipe" works fine for os/360 but fails when running zLinux31/64 Checking file systems... Parallelizing fsck version 1.19a (13-Jul-2000) /dev/dasdb1 has gone too long without being checked, check forced. HHCCP017I CPU0000: Machine check due to host error: Floating point exception PSW=070DE000 C002BF44 INST=B3985012 CFEBR 5,1,2 V:0000058E: Translation exception 0010 Bypass: use default compile options Reproducible: Always Steps to Reproduce: 1.hercules 2.ipl A91 (well ...) 3. I'll keep a copy off these DASD Actual Results: all went well uptill Expected Results: go on, like when I compiled it without optimization
Is the -march=athlon-xp makeing it stuff up? To my understanding your running a emulator for a mainframe and compiling for a athlon-xp. This doesn't make sense to me. Can you be a bit more specific as to what you are doing. Are you emerging something under an emulator? if so what. When does the error occur? during compulation, restarting the emulator, running the application. I realy cannot follow what you are doing.
I compiled/emerge Hercules 3 (with the optimization options etc.) That binary fails when runnin Hercules and emulating zLinux and in the situation I escribed (volume check) with floating point errors. When compiling with less optimization options (/etc/make.conf) it runs well.
So it's actually a compilation option what causes the error, so the emerge is not correct.
Hercules has had problems before with optimization...for the longest time, -O3 on Alpha would make it behave really badly. I'll pass this bug along to the developers. In the meantime, I'll look at the ebuild and make sure it's not doing weird things with the optimization flags; Hercules picks its own optimizations during configure for good reason, and it's not a good idea to override them unless you know what you're doing.
Jay, any news on this one? Has it been sent upstream for instance? Is there still something to be done by Gentoo for this bug? Last activity was over a year ago.
Re-assign wrt Bug 44756... If someone picks this up, please remove killsoft from metadata.xml.
Eh, remove jmaynard is what I meant...
This should be fixed with 3.03. Note that the 3.03 ebuild always uses the standard compile options, instead of allowing the user to specify his own optimization level; Hercules stresses the gcc optimizer, and overriding the default optimization flags leads to weird results - such as are seen here.
(In reply to comment #8) > This should be fixed with 3.03. Note that the 3.03 ebuild always uses the > standard compile options, instead of allowing the user to specify his own >optimization level Minor glitch - there's no 3.03 in the tree... :=)
upgrade to latest versions of stuff