Stephan Witt wrote: > Am 19.01.2016 um 20:10 schrieb Scott Kostyshak <skost...@lyx.org>: >> >> On Tue, Jan 19, 2016 at 08:10:59AM +0100, Stephan Witt wrote: >> >>> BTW, when debugging tex2lyx it complains about missing USERDIR and I >>> have to add -userdir >>> /Users/stephan/git/lyx-build/cmake/2.2.0dev/Testing/.lyx explicitly. Is >>> it passed via environment in ctest runs? Or is it simply not printed? >>> Shouldn't these options passed via command line when calling tex2lyx? >> >> I do not have to do this (unless I set something up automatically >> so long ago that I forgot) so maybe this could be the problem. Kornel or >> Georg will probably have an idea.
I don't need that either. >>> Other possible differences except the OS? :) >>> - version or availability of algorithm2e.sty? >> >> I've tested on latest TeX Live 2015 and also with an older TeX Live 2015 >> (the file says "algorithm2e.sty,v 5.0 2013/01/06 14:40:35“) > > Nearly the same here: > "$Id: algorithm2e.sty,v 5.0 2013/01/06 14:40:35 cfiorio Exp $“ Does not matter. tex2lyx does not read .sty files (it only needs to exist so that the package list for LyX and tex2lyx lists it as available). The warning is expected. >>> - remnants of previous runs somewhere? >> >> I use a clean build directory. > > Me too - I really meant *somewhere* - > perhaps tex2lyx reads some file from outside the build directory. No. It reads from the system LyX dir which is the source dir when running the tests, and from the user dir which is Testing/.lyx when running ctest. BTW, cmake copies stuff from the source dir there, but this is not needed. support::Package is supposed to handle this for non-installed builds. > Tracing Preamble::handle_package I have two calls with: > > package: fontenc, options = T1 > package: inputenc, options = latin1 > > As you already noticed I don’t have the "Ignoring options…“ active. > There are never any options left at the end of handle_package. This is wrong. There should be exactly one of those warnings. > Georg, do you have an idea what I can do to find the explanation for this? Unfortunately I am too busy with other stuff currently, so I could not investigate thoroughly. My wild guess (based on the missing user dir) would be that either something in support::Package is wrong for OS X, or that cmake does not configure the hidden user dir correctly. To rule out the latter, you could run the tests using autotools. Does that work? I'll hopefully have some more time tonight. Georg
