Guenter Milde wrote: > On 2015-11-17, Georg Baum wrote: > >>> However, as the "non-ASCII in preamble" warning is a false positive and >>> LaTeX does not have a problem with "unencodable" characters in a >>> comment, it is no problem to leave these comments in the lyx-document >>> and no wonder that removing them does not solve any failure. > >> Some time ago there has been a huge discussion about treating unencodable >> characters in a comment as errors or not. The result (which was not liked >> by everybody) was that these are treated as errors, and therefore LyX >> behaves exactly as designed. > > The check for unencodable characters does not distinguish comments from > "real" TeX code. > > In ERT, LyX raises an error for every unencodable character. > > In the preamble, unencodable characters trigger a warning, not an error.
Sorry, you are right, my memory was wrong. >> Therefore I think it is a good idea to let the test machinery remove >> comments at least for the XeTeX + TeX fonts tests, until a final decision >> has been made whether we want to make our docs compilable for this exotic >> combination or not. > > I don't agree: don't make tests complicated and opaque only to test this > obscure combination (nobody really needs it). If LyX throws an error, > invert the test. Inverting a test means: "I expect this test to fail, tell me if it does not". If nobody needs the tested combination, don't run tests for it. Running them although nobody cares for the result is only wasting time. I made the suggestion above because it looked to me as if a lot of effort was spent to make the combination XeTeX + TeX fonts work better. I think we really should decide whether we care for this combination or not. If we don't, then we should not spend further time with improving it, and we should stop running tests which test it. If we do care, then we should continue with testing, and if there are volunteers also with improving the code. > (And the good news: in the user preamble, tests with documents where the > only problem is an unencodable character in a comment pass as the warning > is ignored.) This is what I'd expect: Warnings do not cause an error exit code, errors do. However, as I understood Kornel he observed something different. Georg
