From: ludo@gnu.org (Ludovic Courtès)
To: Timothy Sample <samplet@ngyro.com>
Cc: guix-devel@gnu.org
Subject: Re: Graft hooks
Date: Wed, 22 Aug 2018 16:21:48 +0200 [thread overview]
Message-ID: <874lfmh477.fsf@gnu.org> (raw)
In-Reply-To: <87k1ojafqa.fsf@ngyro.com> (Timothy Sample's message of "Tue, 21 Aug 2018 11:42:21 -0400")
Hello Timothy,
Timothy Sample <samplet@ngyro.com> skribis:
> ludo@gnu.org (Ludovic Courtès) writes:
>
>> Hello Timothy,
>>
>> Timothy Sample <samplet@ngyro.com> skribis:
>>
>>> The basic idea would be to add a field (or use a property) to the
>>> package record. Let’s call it “graft-hook”. It would be Scheme code
>>> that gets run after grafting takes place, giving us a chance to patch
>>> special things like checksums. The hook would be passed the list of
>>> files that were been modified during grafting. Then, in the Racket
>>> package for example, I could write a graft-hook that updates the SHA-1
>>> hash of each of the modified source files.
>>>
>>> Since grafting is done at the derivation level, the hook code would have
>>> to be propagated down from the package level. I haven’t looked at all
>>> the details yet, because maybe this is a bad idea and I shouldn’t waste
>>> my time! :) My first impression is that it is not too tricky.
>>>
>>> Are these problems too specialized to deserve a general mechanism like
>>> this? Let me know what you think!
>>
>> I agree that this would be the right thing to do! (I’d really like to
>> do it for GDB as discussed in <https://bugs.gnu.org/19973>.)
>>
>> Package properties would be the right way to make it extensible, but
>> there are complications (notably we’d need to use gexps, but build
>> systems don’t use gexps yet.)
>
> But soon, right? ;)
Well, it’s complicated. :-)
Also, I realized that some things, like the .gnu_debuglink and build-id
hooks, don’t really fit in any package; they’re transverse.
> Here’s a draft patch (it’s mercifully small). I have a few questions
> about it, but if it looks like the right approach, I will clean it up
> and submit it.
>
> Basically, it checks if we are grafting Racket, and then adds some code
> to the build expression to run the hook.
OK. In theory, should it be just for Racket, or should it also be for
Racket libraries (we don’t have any currently AFAIK)?
> Also, is there a preference for patching the files using Guile or using
> an external tool? This patch uses Racket’s “raco setup” command to
> recompile the files and fix the checksums. Unfortunately, it also
> updates timestamps. I’m pretty sure our Racket package is not
> reproducible at the moment, so I didn’t worry about it too much. The
> timestamps could be patched out, though. The reason I shied away from
> writing my own code is that Racket also hashes all the dependencies for
> a bytecode file. This means that the custom code would have to traverse
> the Racket dependency graph to get the checksums right. It is not too
> hard to do so, but it would be a couple hundred lines of code (compared
> to the five or so it took to invoke “raco setup”).
Regarding whether or not to write our own code: let’s do whichever is
more convenient. In this case, using ‘raco setup’ looks like the right
thing to do, given that raco is available in the build environment
anyway (see below); for .gnu_debuglink, I found it nicer (and more fun
:-)) to write a Guile module.
Regarding timestamps: I guess there’s no problem since timestamps are
reset in the store.
Some comments:
> diff --git a/guix/grafts.scm b/guix/grafts.scm
> index d6b0e93e8..88a99312d 100644
> --- a/guix/grafts.scm
> +++ b/guix/grafts.scm
> @@ -75,6 +75,36 @@
> (($ <graft> (? string? item))
> item)))
>
> +(define (fix-racket-checksums store drv system)
> + (define racket-drv
> + (let ((package-derivation (module-ref (resolve-interface '(guix packages))
> + 'package-derivation))
> + (racket (module-ref (resolve-interface '(gnu packages scheme))
> + 'racket)))
> + (package-derivation store racket system #:graft? #f)))
> +
> + (define hook-exp
> + `(lambda (input output mapping)
> + (let ((raco (string-append output "/bin/raco")))
> + ;; Setting PLT_COMPILED_FILE_CHECK to "exists" tells Racket to
> + ;; ignore timestamps when checking if a compiled file is valid.
> + ;; Without it, Racket attempts a complete rebuild of
> + ;; everything.
> + (setenv "PLT_COMPILED_FILE_CHECK" "exists")
> + ;; All of the --no-* flags below keep Racket from making
> + ;; unecessary and unhelpful changes (like rewriting scripts and
> + ;; reverting their shebangs in the process).
> + (invoke raco "setup" "--no-launcher" "--no-install"
> + "--no-post-install" "--no-info-domain" "--no-docs"))))
Since this is used when grafting Racket, I would suggest moving this
graft to the “build side” entirely, similar to what I did in
<https://debbugs.gnu.org/cgi/bugreport.cgi?bug=19973#25>. Probably
you’d just add a single procedure to (guix build graft) and add it to
%graft-hooks.
That procedure could be the same as what you have above, except that
it’d run OUT/bin/raco, if it exists, and do nothing if OUT/bin/raco does
not exist.
WDYT?
Thanks,
Ludo’.
next prev parent reply other threads:[~2018-08-22 14:21 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-08-12 19:49 Graft hooks Timothy Sample
2018-08-13 0:28 ` Christopher Lemmer Webber
2018-08-13 7:19 ` Gábor Boskovits
2018-08-20 9:12 ` Ludovic Courtès
2018-08-21 15:42 ` Timothy Sample
2018-08-22 14:21 ` Ludovic Courtès [this message]
2018-08-22 18:29 ` Timothy Sample
2018-08-24 10:09 ` Ludovic Courtès
2018-08-22 19:15 ` Christopher Lemmer Webber
2018-08-23 7:15 ` Mark H Weaver
2018-08-23 9:54 ` Gábor Boskovits
2018-08-24 10:17 ` Ludovic Courtès
2018-08-24 21:59 ` Mark H Weaver
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=874lfmh477.fsf@gnu.org \
--to=ludo@gnu.org \
--cc=guix-devel@gnu.org \
--cc=samplet@ngyro.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this external index
https://git.savannah.gnu.org/cgit/guix.git
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.