On 2/5/22 21:01, Christine Lemmer-Webber wrote:
Hello!
It's been a while since Guile was my main hacking environment; I've been
returning to it, and one of the nicest things to change about its
ecosystem is the presence of Guile Hall.
I really, really like Guile Hall. A lot! I think it has room to grow
but it fills a clearly missing piece of the Guile ecosystem while doing
it in the best way possible: making itself explicitly compatible with
Guix.
I thought I'd write down some impressions while everything is fresh.
- Its ability to make an autotools-compatible tarball, but without me
needing to think about autotools at all, is a real delight.
- Its test suite stuff is also really nice.
- I found myself surprised that hall.scm is "just data", instead of
taking the more guix'y approach of being code that actually builds a
datastucture. I'm not sure what the goal of this is; there can be
reasons to take that approach but I'm not sure what it is here?
My assumption is that the main reason is so that "hall scan" can
correctly read and then modify and spit out another file, but I'm
not sure.
- What I would actually *really* like would be for the Hall package
definition structure to be a wrapper *around* the Guix package
structure. Then the guix.scm would be really simple: it could just
"peel off" the outer struct. If I wanted to do some smart
modifications of things from there maybe I could. I dunno, something
like this.
- "hall scan" is really cool, but I kind of wish I didn't need to use
it. I'd rather not keep track of any of this stuff at all.
I'd be happy just pointing some code at a directory and say "snarf
up all the .scm files you see therein!"
- I'm currently writing a manual starting in a .org file that's then
converted into a .texi file. I'd prefer if I could find an
entrypoint to insert this into the compilation workflow: a pre-step
to the docs compilation that generates the .texi file *from* my
.org file.
- On that note, it strikes me that Hall's integration with autotools
is great because it means that existing distros don't need to be
aware of Guile *or* Guix. But it also means that Hall probably has
all of the information that it could do all the steps that autoconf
and automake do too. That might be interesting to see that.
Anyway, just some thoughts. Making Guile packages is already much less
intimidating now thanks to Hall's work. Thank you for it!
- Christine
For what it is worth: You do not need to use Hall, if you do not need to run
tests, when installing via Guix or similar things. If you are fine with running
tests before publishing a version of a package to Guix, then you only need Guix
and a package definition. No autotools.
I found all the stuff that is generated by Hall a bit worrying. Lots of
autotools appearing in my repository, which I do not fully understand. Being
forced to use hall scan and so on. Too much magic behind the scenes. Packaging
should be so simple, that I can still easily understand all the things in my
repository, without studying autotools for a week, searching for archaic
variable names and what they to and how everything works. I am not a huge fan of
the JS ecosystem at all, but if we look at the NPM world, there is only one file
that needs to exist, the package.json, and already you have a valid package. Why
can it not be the same simplicity for Guile? And if Guix really requires there
to be a `make` `make check` or whatever, then where is the guide, that tells me
how to hand-craft the Makefile, so that everything works and I don't need extra
tooling to generate files.
I want to be able to edit the guix package definition, point it to the tests
directory, and be done with it. When I use Hall, then I have yet another tool in
the workflow. Again more commands and things to keep in mind. I want less tools
and simplicity. I don't run autotools for running my tests locally, so there is
no theoretical need for that. The maximum I do it a very simple Makefile, which
provides me with a `make test` to run tests or `make` to run the program. It is
just that from the Guix side of things, something would have to be done to be
able to invoke tests properly, without a `make test` command being available,
but knowing where the tests are. They could be declared in the package
definition for example. Why do I need any autotools stuff? Just make Guix call
my SRFI-64 tests.
This is why I decided to not use Hall for guile-fslib. I did not like the
complexity of having all the autotools stuff in my repo, which I don't even
fully understand, just for it becoming a package. Even if I don't need to
manually touch any of those files. I can run tests on my machine before sending
in a patch for a new version and that will be fine.
Hall is probably a great thing, if you have a package, that is not pure Guile,
so that one has to do other things than simply running a Guile command to run
the tests, perhaps to generate documentation or other files. For pure Guile
packages, I still need to see or learn about a good reason to add the complexity
of it to a repository.
Best regards,
Zelphir
--
repositories: https://notabug.org/ZelphirKaltstahl
