Concerning the make problem, this is one thing, probably we can find some dependency error somewhere in the make file, or more likely I speculate that the issue is that with do_java=1 we compile the java sources in two places, in the Makefile and in the t4ht system calls. As make can run several processes in parallel this may create some unwanted timestamp touching. But more generally there is the fact that literate programming changes the time stamps of all the generated java files even though the content might be unchanged, just the documentation thereof. Maybe some more clever dependency based on md5sum checks rather than time stamps could solve that. Another thing is that in general make is not so good for java, because the same file.java can generate several file*.class files, and the number of files generated depends on the file.java content. Maybe delegating java compilation to ant would be preferable.
Anyway, in the meantime here is step4 patch. V. Le sam. 6 juin 2020 à 00:58, Karl Berry <[email protected]> a écrit : > > Here is a fix for the problem that you observed. > > I installed it. Thanks. That fixed the first problem, but not the second > ("pushd: no other directory"). tex4ht-javahelp-xtpipes.tex did not have > the necessary \input{tex4ht-dir} (maybe you had that as a local > change). I committed that too. > > By the way, when the same change is being done in several files, there's > no need to repeat the description in the ChangeLog. Just > * file1, > * file2: change. > suffices. > > I am a bit puzzled why I did not observed the same error. > > I've observed that often errors happen on run N that do not happen on > run N+1. I don't usually try to clean everything before any given run. > > appropriate to disable it in the .tex by default unless do_java is > ... > That would be easy to do through the texh4t-dir.tex that could contain > a switch passed from the makefile. > > If you want to make a patch for that, it's fine with me. I was just > making the minimum change in the shortest possible time :). > > A bigger problem is that even after running make do_java=1 > and everything succeeding, running make do_java=1 again > still causes everything to be remade. That should not happen. > I did not try to trace the dependencies with make -d or whatever, though ... > > Best, > Karl
step4.diff
Description: Binary data
