Guenter Milde wrote: > Generally, I don't think it is a good idea to "massage" the documents > shipping with lyx in the test machinery until the tests pass. > > * If the document can be made more robust (i.e. working with more than the > default output format) without the changes standing in the way for its > main purpose (documentation, templates, examples for LyX users), then > this should be changes in the document itself.
Yes. > * If the document has special needs (packages, ERT, preamble code ...) > that make it uncompilable with some export formats, invert the test or > ignore it - with appropriate comment. There remain enough files to > test. > > If some document in lib/... is important for testing a special > feature/combination/... but fails for another reason, please don't > modify it in the test machinery, but instead convert it to a > *minimal working example* and put in a dedicated export test document > folder. Yes. I started to do this (for language nesting), but getting the needd infrastructure in place is unfortunately more difficult than hoped for. > This is far more easy to handle for other developers than test > passing/failing for reasons that cannot be reproduced when compiling the > same documents "by hand"! Definitely. However, currently we do not have these tests yet, and as an intermediate step the automatic creation of all the different combinations from existing documents is better than nothing IMHO. Once we have the infrastructure for running dedicated small tests, we should definitely focus on them. Georg
