David Bremner <da...@tethera.net> writes: > Adrian Bunk <b...@debian.org> writes: > >>> FTR, the dates when epl started to FTBFS in sid and bookworm are >>> a good match for when emacs 1:28.2+1-9 entered sid and bookworm: >>> https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/epl.html >> >> Correct link: >> https://tests.reproducible-builds.org/debian/history/epl.html > > Some pretty weird behaviour here > > HOME=/n0nexistent emacs -batch -Q -l package \ > --eval "(add-to-list 'package-directory-list > \"/usr/share/emacs/site-lisp/elpa\")" \ > --eval "(add-to-list 'package-directory-list > \"/usr/share/emacs/site-lisp/elpa-src\")" \ > -f package-initialize -L . -L test --eval "(progn (setq > comp-enable-subr-trampolines nil) \ > (load-file \"test/test-helper.el\") (load-file \"test/epl-test.el\"))" \ > -l test/epl-test.el --eval \(ert-run-tests-batch-and-exit\) > > fails, but replacing /n0nexistent with /nonexistent passes. I hope this is > not a sign that someone has special cased /nonexistent, but I fear otherwise.
This weird behaviour is a consequence of emacs' ill-advised special-casing of "HOME==/nonexistent" in startup.el (around line 550). if HOME==/nonexistent, then emacs tries to provide a temporary directory for native compilation output. Why this is needed is a different question.