Untrusem <[email protected]> writes:

> So I am updating rust to rust-1.91, where the 'build phase succeeded but
> the 'install failed, so if one would want to test new changes in the
> 'install phase, they would have to go through the rebuild phase once
> again which in case would be over 1 and half hour for me.
>
> So is there a way to use the last build's contents so we don't have to
> rebuild everytime we want to test a change basically restarting the
> build.

This would require determining which package definition fields influence
which build phase, and changing a build to be a chain of derivations
rather than just a single derivation (if I'm not mistaken) - I think
that that would be theoretically possible, but extremely unpractical.

Instead, I suggest running `guix build -K` and manually examining the
build directory - you can then manually run the commands that failed in
the build dir after making some adjustments, hopefully allowing you to
figure out a correct build recipe more quickly.

I think tooling to manually and imperatively run individual build phases
of a package/build system in a specific directory might make the process
of poking around in a build dir more effective, and I (baselesly) think
that such tooling would be relatively easy to implement

> As kestrelwx mentioned in irc that this would use a lot of disk space to
> keep each phase, so maybe an option to specify if we want to keep a
> specific build phase would be a workaround?

Indeed, keeping around *all* build phases would make the store balloon

> Futurile told me there was some discussion about this last year but they
> can't seem to find the exact patch series.

Reply via email to