Marius Bakke writes: > "Taylan Ulrich Bayırlı/Kammer\"" writes: > >> I struggled for days with the following bug and ultimately decided to >> disable tests for 1.73 to sidestep it entirely; tests should be >> re-enabled with the next release. >> >> https://github.com/DanBloomberg/leptonica/issues/224 > > OK! I'm not familiar with this software, are you sure the substitutions > are no longer required? The pkg-config file is fine as there is no > matching substitution in the current version, but I did not investigate > the config.h timestamp trick, or the failing tests. > > If they are required for tests to pass on current git, I think they > should be kept so the next person don't have to reinvent them. I would > also keep gnuplot as a native-input, but that can be fixed up before > committing if you confirm that the other phases are no longer needed. > > Thank you! You know what, Leptonica 1.74.0 has been released yesterday, so here's a new patch. :-D But yes, the source snippet / patch for the .pc file isn't necessary anymore for reasons not really clear to me. (The two libs still need to be propagated and Tesseract has to link against them; maybe Tesseract special-handles this and other dependents of Leptonica could still benefit from a patched .pc file, but so far we don't have any other dependents of Leptonica...) The config.h hack probably wasn't necessary anymore since 1.72; I had introduced it with 1.71 where the issue existed and it didn't get removed upon the 1.72 update. And all tests pass, without having to disable parallel tests either. The 1.74.0 tag on GitHub corresponds to a state in which ./configure doesn't exist; the autobuild script needs to be run. So here's the new patch.