On Sat, Aug 21, 2010 at 7:47 AM, Dr. David Kirkby <david.kir...@onetel.net> wrote: > Setting SAGE_CHECK=yes forces the test suites of some packages to run. > Clearly that slows the build time - in some cases very significantly. I > think with Pari, it takes about 10x as long to run the test suite as it does > to just build the code. > > But for some programs, the time need to run the test suite is very small > indeed. (Many take less than 10 seconds on my Sun Ultra 27). > > Would it not be sensible to always run the test suites on packages where > they run quickly and exit if there are errors? > > I'd suggest as a cut-off, using 30 seconds on sage.math. In other words, if > a test suite can be run in under 30 seconds on sage.math, then run it every > time Sage is built. If it takes longer than that, then only run it from > spkg-check. > > I doubt this would have much impact on the time to build Sage, but it may > improve the quality. The reasons I believe it will have minimal impact is > > * Only a small number of packages have test suites. (About 20% to date). > > * Of those that do have test suites, about 50% take less than 10 seconds to > run on my Sun Ultra 27 (probably a bit longer on sage.math) > > * Several of the test suites would fall outside the 30 second cutoff, so > would not get run anyway. > > Despite the tile of the ticket, if you ignore anything other than the first > two columns here, you will get a list of what packages have spkg-check > files.
+1 to running all short test suites on build if they're small enough (e.g. less than 30 seconds, or, say, 25% of the build time). That is assuming they're expected to pass, unlike some (scipy?). - Robert -- To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an email to sage-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
