Nicola Musatti said: (by the date of Wed, 02 May 2007 23:40:57 +0200)
Sohail Somani wrote: [...] Considering the 1.34 regression tests, which only cover officially supported platforms, between gcc and msvc there are 16 variants. You may decide to cover only a subset, but you must test all the others to ensure you didn't break anything. I wish I could afford to run so many regression tests each night.
I still think that a revived bug traking tool, where people who care post warning clearing patches, is more viable.
IMHO a following statement should be put into 1.34 release notes. And perhaps even on the boost front page in news regardin the release: "We are working to remove unnecessary warnings caused by boost headers during compilation. This task is not easy due to big number of supported compilers and removing a warning in one of them causes a warning in the other compiler. Moreover the regression tests are unable to catch all possible warnings - they mostly come up during regular usage. We kindly ask our users to submit "remove warning" patches to the bugtracker. We hope that by this action the next release will have a much smaller amount of warnings" -- Janek Kozicki |