On Thu, Sep 03, 2026 at 04:17:16PM +0200, Bruno Haible wrote: > Patrice Dumas wrote: > > > > I am not sure about what to do. One option would be to add a > > > > Makefile.am in tta/perl/t and in tta/perl/t/results and a Makefile.am > > > > in each of the tta/perl/t/results subdirs and generate the list of lies > > > > for each of those directories. > > Will this solve the problem, or will the problem reappear when there are > 3000 or 4000 files?
It should solve the problem at least for such a doubling, since we have 40 directory, as the number of files per directory should be smaller. Also we can split a test in two when it becomes too big. We actually did it before (sectioning split in sectioning and moresectioning), maybe because of a similar issue of too many arguments for a testing command, or because it is better to have similar run time in each t/*.t file for paralellization, so in each directory. > I have a similar thing in GNU gettext. Quite easy: > > dist-hook: dist-many-files > > dist-many-files: > $(MKDIR_P) $(distdir)/ > for file in `cat $(srcdir)/many-files.txt`; do \ > if test -f $$file; then d=.; else d=$(srcdir); fi; \ > cp -p $$d/$$file $(distdir)/$$file || exit 1; \ > done Wouldn't "in `cat $(srcdir)/many-files.txt`" as for argument suffer from the same length issue? If it works, that would indeed be good. -- Pat
