On Thursday, May 03 2012, Jan Kratochvil wrote: > Hi Sergio,
Hi Jan. > On Fri, 04 May 2012 01:25:04 +0200, Sergio Durigan Junior wrote: >> I've been wanting to setup a regression tester for GDB, which would run >> daily on x86_64 initially (the best time to run this task would also be >> something to be decided here). > > the test should cover wide range of distributions. Patch submitters run > testsuite on their own box, just duplicating the submitters work is not > useful, the goal is to test more thoroughly. Proofs why thorough testing is > needed just From the top of my head: Thanks for the message, I appreciate your examples and your rationale. I know that: a) it is difficult to test every distro/arch/whatever possible combinations b) there are some bugs very specific to some distro/arch/whatever, which are hard to catch c) GCC Farm won't (and doesn't need to) provide all the necessary combinations above However, as a starter, I thought it would be useful to set up a regression tester on GCC Farm (like a "public" regression tester), even if it covers only a very small set of combinations of distro/arch/whatever. As we have talked before, I don't have enough machine power to set up a full regression tester as you did. I also miss a public archive of tests where I can find information about regressions (even if it's daily, and not per-commit). > Just x86_64 also is not enough: > http://sourceware.org/ml/gdb-patches/2012-04/msg00896.html > - crash only on i686 hosts My idea was to initially use only x86_64, but expand (if possible, of course) to other architectures. Also, I think it should be fairly easy to setup a tester for both x86_64 and x86 on the same machine. > So I believe there should be some sort of virtualization (I use only mock now) > as I doubt there is enough hardware boxes each with different OS to cover all > the most popular distros. > > > Then also if it runs only daily (and not per commit) there is the manual work > of bisecting each regression / failing new testcase by hand. Yes, as I explained above, initially the idea would be to run the test only daily, also in order to avoid overloading the farm machines. Well, of course my message was only a question and I won't complain if people say that it's useless to setup such a regression tester on GCC Farm. But since I miss a central point where I (and anybody else) can find public information about possible regressions on GDB, I decided it was worth trying. Thanks, -- Sergio _______________________________________________ Gcc-cfarm-users mailing list Gcc-cfarm-users@gna.org https://mail.gna.org/listinfo/gcc-cfarm-users
