bug#81517: 31.0.90; executable-find on remote buffer fails
Sean Whitton <[email protected]> Thu, 30 Jul 2026 11:48:15 +0100
| Newsgroups | gmane.emacs.bugs |
|---|---|
| Message-ID | <[email protected]> |
Eli Zaretskii [30/Jul 7:29am +03] wrote: >> Cc: [email protected] >> Date: Wed, 29 Jul 2026 18:38:54 +0200 >> From: Michael Albinus via "Bug reports for GNU Emacs, >> the Swiss army knife of text editors" <[email protected]> >> >> Shun-ichi TAHARA <[email protected]> writes: >> >> Hi, >> >> > After Emacs31.0.90 with Windows, executable-find on remote buffer >> > returns nil. Then, for example, save-buffers on remote file using tramp >> > fails because '(executable-find "pscp")' returns nil; I use "pscp.exe" >> > derivered from PuTTY and it is expected that the exe file is found by >> > executable-find. >> > >> > I found that tramp added '(exec-suffixes . (""))' to tramp-connection- >> > local-default-system-variables at tramp-integration.el since Emacs31. >> > So 'exec-suffixes' is always changed to that value on remote buffer. >> >> Yes, this is an undesired side effect of bug#78886. >> >> > It harms because suffixes are disabled even when local executables are >> > required; "pscp.exe" installed in local file systems must be found then. >> > >> > I suppose it might be a bug of tramp. >> > Emacs-31.0.91 has the same problem, also the same with git HEAD. >> >> It has been fixed with bug#81280. The fix is in master, not in the >> emacs-31 branch. It covers also call-process and make-process. Shall we >> backport it to Emacs 31? Eli, Sean? > > Ouch! I guess we must, since this whole exec-suffixes business is new > in Emacs 31. > > Sean? We are discussing backporting the change in dcdd76f89e7, right? I think that is okay to backport -- /if/ it introduces problems, they are likely to be no worse than what it fixes. -- Sean Whitton