2012/2/1 Michael Stahl <mst...@redhat.com>: > On 01/02/12 12:42, Stephan Bergmann wrote: >> On 02/01/2012 12:01 PM, Michael Stahl wrote: >>> actually regressions should be easy to prevent here: just build 2 full >>> trees with all options, e.g. enable all extensions, help, all languages; >>> then diff the files produced by the tools. >> >> In principle at least (and I'll surely do tests along these lines). In >> practice however, things are a little harder. Complete input data >> (--with-lang=all) is huge; what parts exactly of the build (*/$INPATH/*) >> to copy aside for later comparison is not always glaringly evident; >> doing changes incrementally would conflict with other people's commits >> invalidating the comparison data saved upfront. > > hence "build 2 full trees" > > yes, that will take lots of disk space.
New tool does not have to produce identical result. It must produce syntactically equivalent result. If we reproduced all tab/space/linefeed uglyness in generated xml/src files with new code, then new code would be as ugly as old. :) So plain old diff is not the right tool for this IMHO. Andras _______________________________________________ LibreOffice mailing list LibreOffice@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/libreoffice
