[bug#50227] [PATCH 0/3] go-build-system and GOPATH improvements

[Marius Bakke <marius@gnu.org>]

[-- Attachment #1: Type: text/plain, Size: 8566 bytes --]

Hi Sarah,

Sarah Morgensen <iskarian@mgsn.dev> skriver:

> Hello Marius,
>
> (Apologies in advance for the length of this treatise! I did not have
> the time to be concise.)

No hurries, the insightful feedback is much appreciated.

> Marius Bakke <marius@gnu.org> writes:
>
>> These patches adjust the Go build system to use Guix's regular
>> native-search-paths mechanism instead of ad-hoc GOPATH trickery.
>
> I have been working on overhauling the Go build system behind the
> scenes; I expected to have a patch ready last week, but I fell down the
> "modules" rabbit hole after learning GOPATH is expected to be deprecated
> as soon as 1.18.  Sorry for the duplicated work!  (I also have a Go 1.17
> ready to launch, but I've been attempting to nail down Go build system
> changes first so I didn't introduce anything incompatible.)

That is excellent news, thank you!  I briefly looked into "modules"
while researching this, but realized that the rabbit hole would be too
deep for me.  I did not notice that GOPATH was being deprecated however.

> In any case, I hadn't thought of using search paths, that's quite
> clever!  I like it.
>
> Before falling down the "modules" rabbit-hole, here is what was working
> for my GOPATH-based Go build system:
>
> 1. Install source in "out/share/go/src" rather than "out/src", and then
> simply create a directory union of "/share/go/src" from inputs.  This
> avoids accidentally including non-go packages with a "/src" directory.
> If you did this, then you could make the GOPATH search path be
> "/share/go/..".
>
> 2. Split the GOPATH as you have done (except with only two components;
> one for the package we're building and one for the union in 1).

Installing to $out/share/go is much nicer indeed.

> 3. Reuse build artifacts by copying the $GOPATH[0]/pkg directory to
> "out/share/go/pkg" in the install phase.  They will be transparently
> used since they will be in GOPATH.  (You can use "out/lib/go/pkg", but
> you must set 'strip-directories' to avoid stripping Go archives, and
> then include "/lib/go/.." in the GOPATH search path.)  However, "go
> install" will eventually deprecate installing archives [0], perhaps even
> before GOPATH is deprecated.

Right.  The churn in Go's build tooling is surprising, I would expect it
to slow down at this point!  I suppose we can ignore the "pkg" artifacts
for now, or keep them in "share/go" as a stop-gap, since they will be
deprecated soon.

> 4. Use -trimpath instead of remove-go-references, as you did.  Also, to
> avoid rebuilding the standard library with '-trimpath' for every package
> (since the Go build cache does not persist between build environments):
>
>   a) modify the Go package to build standard libraries with -trimpath,
>      which would unfortunately mean most users of the Go package would
>      find that ~180MB of space wasted; or
>   
>   b) build a '-trimpath' version of the standard library separately and
>      use it with '-pkgdir' (which would prevent #3 from working) or by
>      building a directory union of Go and Go-std-library-with-trimpath
>      and setting GOROOT=/path/to/union.
>
> Personally, I'm partial to a), along with removing the pre-compiled
> standard library from the Go package since it ends up recompiled more
> often than not, is very fast to recompile, and it will eventually no
> longer be distributed or used by Go [0].

Removing the compiled libraries sounds fine to me.  I suppose we'll
still need -trimpath for executables ("main.go")?

>> The context is that I needed to hack on a Go package, and was somewhat
>> surprised that my usual workflow of "guix environment PKG" did not work.
>>
>> It still does not work "out of the box", but these patches bring it a
>> step further.  Now "all" that is needed is to:
>>
>>   $ cd ~/src/go-foo
>>   $ guix environment go-example-com-foo
>>   $ MYGOPATH="$HOME/tmp/go"
>>   $ NAMESPACE="$MYGOPATH/src/example.com/foo"
>>   $ mkdir -p $(dirname $NAMESPACE)
>>   $ ln -s $PWD $NAMESPACE    # or git worktree add $NAMESPACE
>>   $ export GOPATH="$MYGOPATH:$GOPATH"
>>   $ go build                 # no 'go get' necessary!
>
> Interesting.  I hadn't thought of the use-case for actually hacking on go
> packages like this!  I'll have to think of how modules-mode can be made
> to work with this.

It's not a very important feature, and I'm happy to scratch that itch
again once gomodules are first-class.  :-)

> (A digression: the current issue with fully implementing module-aware
> mode is that Go really wants a specific version for each dependency.  If
> we just populate the module cache with the versions we have, it will
> inevitable complain when a package we try to build wants a version we do
> not have.  I see a few solutions:
>
> 1. Put all dependencies in the module cache, and rewrite the main
> module's go.mod (that is, add replace directives) to replace all
> dependencies with the versions we have.
>
> 2. Rewrite the go.mod to replace all dependencies with the local
> filesystem path of the versions we have.
>
> 3. Put all dependencies in the vendor/ directory, and use -mod=vendor.
> Any pre-existing vendor directory must be handled properly.
>
> These three solutions fail to allow re-using the build cache (and
> therefore build artifacts), because Go computes the build cache keys
> differently for main and non-main modules.  Building in Go is generally
> fast, so we probably shouldn't compromise much to enable reusing the
> build cache, but a few ideas for doing so:
>
> 4. Set up a dummy go.mod out of the source tree, which 'replace's all
> dependencies AND the module we're building like in 1) or 2).  This may
> have to account for replace directives in the go.mod of the module we're
> building, though.
>
> 5. Put the module we're building in the module cache, and build it with
> "go install module@version".  The same caveat as in 4) applies, as well
> as that "go install module@version" only works for main packages (that
> is, packages which produce an executable).)

Thank you for this analysis.  The vendoring option is compelling, if it
does not require patching the go.mod files, and can work also for
packages where unbundling is not feasible (or downstream channels with
less strict packaging policies).

For reusing build artifacts, perhaps we can piggy-back on whatever is
implemented for Bazel as mentioned in [0].

>> I don't know how feasible it is to avoid making a local directory and
>> symlinking the project to the expected namespace.  Still a complete Go
>> newbie, but this approach feels more natural and idiomatic Guix-wise.
>
> My intuition is that if you're working in GOPATH-mode, you already have
> a ~/go/src directory or similar, and your project is probably under
> ~/go/src/my-project. Then, in order to hack on it Guix-like, you would
>
>   $ cd ~/go/src/my-project
>   $ guix environment go-github-com-me-my-project
>   $ export GOPATH=~/go:$GOPATH
>   $ go build

That is much easier.  You can tell I never hacked on Go before.  :-)

> I'm not sure what a similar idiom for Guix-like hacking in module-aware
> mode would be; we'd have to set GOMODCACHE or something, but it would be
> very easy for Go to overwrite (or fail to overwrite) things without
> GOPROXY=off.  Alternatively, if we make a "full" go proxy directory
> layout, we can do
>
>   GOPROXY=file://path/to/gomodcache
>
> or even a search path like
>
>   GOPROXY=file:///gnu/store/p1/gomodcache|file://gnu/store/p2/gomodcache
>
> though I'm not sure how well that would scale w.r.t. number of packages.
>
> Both of these GOPROXY methods have the advantage over setting GOMODCACHE
> that the user could modify GOPROXY to include the default proxy, and
> would still be able to get packages and versions not packaged by Guix.
>
> I suppose there's no reason we couldn't set both GOPATH and
> e.g. GOPROXY.

GOPROXY seems like a great middle ground for local development.

Given the conflicting work here, what do you think we should do?  I'm
happy to scrap this PR as it was largely an exercise to learn
go-build-system, in addition to scratching a very minor itch.

Is the reduced complexity worth it while waiting for the gomodules
rewrite, and if so, are there parts that can be merged with your work
such as using $out/share/go?

Let me know if I can be of assistance.  :-)

> [0] https://github.com/golang/go/issues/47257

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 247 bytes --]