[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[bug#68163] [PATCH] gnu: Prevent stale cache use when `%package-module-p
|
From: |
antlers |
|
Subject: |
[bug#68163] [PATCH] gnu: Prevent stale cache use when `%package-module-path' is parameterized. |
|
Date: |
Wed, 17 Jan 2024 00:32:39 -0800 |
|
User-agent: |
Cyrus-JMAP/3.9.0-alpha0-1374-gc37f3abe3d-fm-20240102.001-gc37f3abe |
Updates:
> I don't know if a.) there's any cause to be concerned about the
> performance of this tweak, or b.) that you care to support the
> behavior at all, but I hope I've made my own use-case clear and the
> patch is there if you see fit c:
a.) I glued hyperfine to a pair of guix repl endpoints, scaled up until I hit
segfaults, no difference
b.) Maybe `specification->package' just has some particular notion of what's
"current", like "current-channels" does, in which case we can rule out
complacence with the parameter as a non-feature-- but I certainly isn't the
same notion.
c.) I had the bright idea to set GUIX_PACKAGE_PATH at build time, first on the
command-line and then in the channel modules, but it didn't work out-- felt
silly for a minute there. Hey, how do third party channels usually resolve
`search-path'? Isn't that the same thing, packages and patches in the same
repo? Or are they just expected to use `local-file' and relative paths? I'll
look into this another time.
If B or C had come to fruition I'd have closed the issue with 'em, and
something like B is up for interpretation-- I don't mind either way (happy to
carry the patch on my own). But if I enjoy these instances of syntax-sugar in
my own channels (which I do :p), and if the Guix repo utilizes + is cleaner for
them as well, then I figure it'd be a shame (barring a solution like C) to
interpret the implementation as one which excludes other channels.