Hello, On Thu 03 Aug 2023 at 07:40am +02, Aymeric Agon-Rambosson wrote:
> tl;dr : The changes introduced by Andrea Corallo following said discussion > have made patches 12 and 13 (the rebased version of Lars Ingebrigsten commits) > useless, so we should drop them. > > In fact, I thought that this was the unanimous conclusion of the discussion, > hence my mail to Sean from yesterday where I enquired as to why patches 12 and > 13 were still present. > > For context, the earlier version of patches 12 and 13 were introduced to solve > a series of FTBFS bugs we had with various elpa-* packages (see #1021842, for > instance). Those bugs happened roughly like so : > > 1. Whenever a primitive (a function available from elisp, but implemented in > C) is advised, it is necessary to compile a trampoline to allow to jump > dynamically to the native-compiled definition of the advice, which could be > anywhere on the user's system. > 2. Said trampoline needs to be output to a file. > 3. The job of function comp--trampoline-abs-filename is to find a directory in > which to store this file, and return it. > 4. This search was made from a list of candidates (think > /usr/lib/emacs/29.1/native-lisp or ~/.emacs.d/eln-cache). > 5. If no such candidate was available, either because they were unwritable > (the former) or because they did not exist (the latter, in our build > environments), the function errored out, which caused our bug. Thank you for the investigation. My conception of the environment variable is a bit broader than just this particular bug: I think that it is a good idea to be able to switch off native compilation in the way this environment variable permits. So I am inclined to keep the patches. -- Sean Whitton
signature.asc
Description: PGP signature
