On 05/30/2012 03:53 PM, Stefan Knorr (Astron) wrote:
first: Stephan, I'm very sorry for breaking your tinderbox(es) for days and not noticing it. (I just assumed my 2-byte change of a commit was innocent enough. I'll do make check in the future as you advised.)
No problem. And picking up using "make check" routinely is highly appreciated.
With some dumbluck and [1], I've managed to get it down to 1 failing test. My approach so far was to just unconditionally add a sheet (in the case of the XMLStylesExporter and XMLSettingsExporter I just made sure these tests actually ran on Sheet 1 (=0), not Sheet 2 (=1)). Since I guess my approach is less than optimal, do tell me if I need to wrap stuff in try/catch or so. Or, if I should ask LibO how many sheets there are before creating a new one. Or something else.
The approach looks rather sane to me (and apparently to Kendy also, who already pushed). In general, the qadevOOo based tests are rather a mess, and nobody loves them. However, they proved useful in finding regressions over and over again, so we should see to keep them working until newer, better tests replace their code coverage (and even go beyond).
About the last failing test, sc_unoapi/ScModelObj, I am unsure where it needs the second sheet, so I didn't add anything yet. Advice appreciated.
I just extended your approach of explicitly inserting an additional sheet to qadevOOo/tests/java/mod/_sc/ScModelObj.java (see <http://cgit.freedesktop.org/libreoffice/core/commit/?id=0210ec32f779bb2098657eb4ad69cca7a60835dc> "Fix one more test that implicitly assumes more than one Calc sheet") and it appears to work well.
So I also re-reverted the "Use one sheet as default in Calc" feature (see <http://cgit.freedesktop.org/libreoffice/core/commit/?id=71056d59de59561ee1d74a3c2d9d66a223422c7e> "Revert 'Revert 'fdo#50183 Use one sheet as default in Calc''").
Stephan _______________________________________________ LibreOffice mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/libreoffice
