Mark H Weaver <m...@netris.org> skribis: > l...@gnu.org (Ludovic Courtès) writes: > >> Mark H Weaver <m...@netris.org> skribis: >> >>> l...@gnu.org (Ludovic Courtès) writes: >>>> Honestly, I wouldn’t worry about the propagation of $GUILE_LOAD_PATH & >>>> co. to subprocesses, because we know there’s none anyway. >>> >>> That policy will lead to future where libguile-using programs break in >>> random ways when they happen to be subprocesses of each other. >> >> I agree in general with your feeling. >> >> However, in that case, we know that these command-line tools are just >> wrappers around our Scheme APIs, and that they won’t ever launch any >> program (programs are a thing of the past; procedures are the future). >> So it just seemed safe to me to do that in this particular case. >> >> What do you think? > > Ah, okay, I didn't realize that. When you said "we know there's none > anyway", I thought you meant "no subprocesses that use Guile", but I > guess you meant "no subprocesses at all".
Exactly. > I guess guix-daemon is the only one with subprocesses, and by the time > that's written in Guile hopefully Guile will have a command-line option > to augment %load-compiled-path. Actually, guix-daemon spawns processes written in Guile, such as list-runtime-roots and hopefully soon a “binary substituter”, but these should be simple stand-alone programs. > In that case, I withdraw my proposal. I'll make a new patch. Thanks! Ludo’.
