Description of problem: On fedora sparc build fail. Version-Release number of selected component (if applicable): hdf5-1.6.6-3 How reproducible: on a sparc machine with fedora port: rpmbuild -bb hdf5.spec --target=sparc Actual results: 1) total crash of H5detect fortran autocionfig helper routine. 2) incorect type declaration specific for sparc lead to compile errors. Expected results: 1) build smooth. 2) pass all tests. Additional info: Attacehd 2 patches. First one patch: It fix proper SEGV signal return in case of bus-error or seg-fault. That mean that fortran part is on his own autodetect for H5detect failed, but it must bail out with 0 error result during compile time, and no header template is generated. This patch should fix any other arch wich cannot relay on H5detect tool during compile time, so i think it would be smart to drag it in. Second one patch: It declare some proper custom types for sparc specific so fortran compilation should go smooth.
Created attachment 290902 [details] trivial patch to return proper signal from H5detect on some arches
Created attachment 290903 [details] fix fortran compile for sparc. Package compile and package self tests went smooth on my u10 sparc box ;-)
Orion, you might want to file booth for of hdf/hdf5 ones upstream ? Or let me please know if you want to do it for you, that will be no problem. ~cristian
BTW, first one patch make sense for rpmbuild wich is pick if a tool/script return bad signal to rpmbuild, whole rpmbuild bail out. i think upstream folks like to do evrything manualy so they dont care of signal return, or something ike that could be my explanation of this lattent bug.
s/pick/pikky/
Thanks for the patches. Applied in hdf5-1.6.6-4.fc9.