On Thu, Mar 10, 2005 at 10:36:49AM -0600, Chris Lattner wrote:> BTW, I added a line to the shell script we use to update the > NightlyTest.pl file from CVS before running it.I thought everyone already did that :) -- Misha Brukman :: http://misha.brukman.net :: http://llvm.cs.uiuc.edu
On Thu, 10 Mar 2005, Vladimir Merzliakov wrote:>> Do to popular request, I just set up a new mailing list to capture the >> nightly tester output from the various testers running 'round the world. >> The new list is available here: >> http://mail.cs.uiuc.edu/mailman/listinfo/llvm-testresults >> >> If you're running a nightly tester, please update your crontab to send mail >> to the llvm-testresults at cs.uiuc.edu list (potentially in addition to >> yourself) so the summaries from the tester will get archived. Internally >> our crontab line looks like this (e.g., for the ppc tester): >> > I update script for http://npt.cc.rsu.ru/testresults-X86-FreeBSD/ > nigthtester.Thanks!> I think can be vary helpful if LLVM and (more important) LLVM CFE have > daily versions like GCC and print its in logs and night test result > pages. LLVM CFE not rebuild at each night test run and some tests can > fail and not fail in dependency from current LLVM CFE.I think the best way to deal with this in the short term is to output the date stamp of the "cc1" program in llvm-gcc. If we output the date in the Dejagnu test results section, it would be easy to see that for a CFE of date N, all C{++}Frontend tests dated > N can be expected to fail. Eventually, the CFE will be split off to run as a proper project, and the tests will be distributed with it. I don't think anyone is planning on working on this though. BTW, I added a line to the shell script we use to update the NightlyTest.pl file from CVS before running it. This might be a useful addition to your script (to avoid making you update it when the .pl file changes). Thanks! -Chris -- http://nondot.org/sabre/ http://llvm.cs.uiuc.edu/
>> BTW, I added a line to the shell script we use to update the >> NightlyTest.pl file from CVS before running it.Ok. I add 'cvs update' command for NightlyTest.pl file to script> I thought everyone already did that :)Now maybe everyone :) Vladimir
>> I think can be vary helpful if LLVM and (more important) LLVM CFE have >> daily versions like GCC and print its in logs and night test result >> pages. LLVM CFE not rebuild at each night test run and some tests can >> fail and not fail in dependency from current LLVM CFE. > > I think the best way to deal with this in the short term is to output the > date stamp of the "cc1" program in llvm-gcc. If we output the date in the > Dejagnu test results section, it would be easy to see that for a CFE of > date N, all C{++}Frontend tests dated > N can be expected to fail.I prepare patch for NightlyTest.pl and NightlyTestTemplate.html to do this without modifing llvm-gcc. Also patch for NightlyTest.pl fix problem with "Number of object files compiled: 0 " in all nightly testers results. In current *-Build-Log.txt files compilation logged like "llvm[3]: Compiling CloneFunction.cpp for Debug build" This set 'Number of object files compiled' to zero with current grep pattern '^Compiling'. Vladimir -------------- next part -------------- A non-text attachment was scrubbed... Name: NightlyTestTemplate.html.patch Type: application/octet-stream Size: 354 bytes Desc: not available URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20050311/a4d91500/attachment.obj> -------------- next part -------------- A non-text attachment was scrubbed... Name: NightlyTest.pl.patch Type: application/octet-stream Size: 882 bytes Desc: not available URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20050311/a4d91500/attachment-0001.obj>