On Sun, May 07, 2006 at 05:00:23PM +0200, Uwe Stöhr wrote: > Personally I use a test procedure for every new release of my > LyXWinInstaller. Derived from this here's my proposal for new releases > of LyX:
Some time ago I started a lyx-tests CVS repository for precisely this regression-testing purpose. Unfortunately it never caught on with anybody else. I'd strongly encourage you to get write access to the module (assuming it got converted) and start committing test documents. > - Test scrolling and editing (copy/paste, etc.) with a long and big > document like a PHD thesis. I believe we knew about all the problems here, but released anyway. > I mean that this could have avoided bugs we had in LyX 1.4, for example > the math bugs I described in the thread "major math bugs", the slowness > in large documents when scrolling and editing, the spellcheck and change > tracking issues, etc. . Several of these were certainly known about. It may be feasible to create some automated tested for LyX rendering issues, both in the program and in the output. The notion would be based around using a well-known setup on a machine under lyx.org so fonts etc. don't change, and grabbing dumps of the root document window. It wouldn't work for lots of things but /anything/ we can automatically test would be good. tex2lyx regression testing could be easily automated. regards john
