On 15/12/11 11:34, Stephan Bergmann wrote: > On 12/15/2011 10:58 AM, Tomáš Chvátal wrote: >> So unless you can look at one of those traces and tell me that the >> build should be stopped because "it is obvious error in the code (not >> in the test, the app code) that needs to be fixed" I would want you > > Hard to tell without a backtrace.
our new so-called unit tests are not as reliable as i would like: i have seen this in build from scratch myself, sometimes the hwpfilter test crashes, and it never happens when building only that module. i guess the reason is probably that it needs some library loaded via UNO, and concurrently with the test another process overwrites that library (have seen exactly that with a test in sw and fixed it), making the test very unhappy. the problem i see is that most of our CppUnit test are not really unit tests but system level tests, and really need most of a full office with UNO stuff and configuration to run. figuring out which UNO components are needed to run the test is a real pain and so of course lots of dependencies are missing. we wouldn't have this problem if these were subsequenttests; it would even be possible to run them when rebuilding a single module with something like "make build && make subsequenttests" (perhaps "make check" in a module should do that?), just not on the build from scratch. _______________________________________________ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice