Check the docs as part of check-stageN

[james-w]

This addresses #31587 by adding check-stage2-docs as a dependency
of make check.

With this change make check output shows

LD_LIBRARY_PATH=/home/jw2328/devel/rust/x86_64-unknown-linux-gnu/stage2/lib:/home/jw2328/devel/rust/x86_64-unknown-linux-gnu/llvm/Release/lib:$LD_LIBRARY_PATH x86_64-unknown-linux-gnu/stage2/bin/rustdoc --test doc/error-index.md

running 453 tests
test Rust_Compiler_Error_Index_0 ... ok
test Rust_Compiler_Error_Index_100 ... ok
test Rust_Compiler_Error_Index_1 ... ok
test Rust_Compiler_Error_Index_10 ... ok
test Rust_Compiler_Error_Index_103 ... ok
test Rust_Compiler_Error_Index_101 ... ok
test Rust_Compiler_Error_Index_105 ... ok
test Rust_Compiler_Error_Index_104 ... ok
...

I'm not entirely sure this is the right fix, we may instead want to depend
on a slightly more specific target, but I'm not sure what the
expected behaviour of each target is. Running all the docs tests
on make check seems perfectly reasonable, and the dependencies
should mean there is no duplicated work.

If you do suggest a change I would appreciate it if you could also
let me know if there's a trick to avoid the process from deleting
all the build artefacts if the makefiles are changed. A simple tweak
to this rule means that I have to rebuild all of rust to test it. If there's
a way to avoid running get-snapshot then I would like to know,
because NO_REBUILD=1 doesn't seem to be it.

This is my first PR for rust, so I likely missed some simple things
(commit message template, NEWS entry?) Please let me know
what they are and I will happily adjust my submission.

Thanks,

James

[rust-highfive]

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @brson (or someone else) soon.

If any changes to this PR are deemed necessary, please add them as extra commits. This ensures that the reviewer can see what has changed since they last reviewed the code. Due to the way GitHub handles out-of-date commits, this should also make it reasonably obvious what issues have or haven't been addressed. Large or tricky changes may require several passes of review and changes.

Please see the contribution instructions for more information.

[james-w]

Looking again it looks like check-notidy probably wants this adding as well right? Do you
know why check is not defined as

check: check-notidy tidy

?

[james-w]

To answer my own question, avoiding a full rebuild on Makefile changes can
be done with

$ make check NO_REBUILD=1 NO_MKFILE_DEPS=1

[alexcrichton]

Perhaps this might be most appropriate behind the check-stage2 rule rather than check? That way make check-stage1 would also run the stage1 doctests

[james-w]

Hi, that's a good idea, I've updated the branch to add check-stageN-docs to
the deps for check-stageN, so each stage will run the doc tests.

[petrochenkov]

IIRC, something was wrong with check-stage1-docs.
Could you verify that it works as expected and doesn't build anything excessive (stage2 for example)?

[james-w]

Ah, it looks like it might build the error generator from stage2.

[alexcrichton]

In the case of make check-stage1, does it then build a bunch of stage2 artifacts? It may just be a bug in where the error index generator comes from (hardwired to stage2 when it maybe should be a parameter)

[james-w]

https://gist.github.com/james-w/6d88ee407f2fd180c04d is the output of make check-stage1 with this change. It looks like it is building some stage2 libraries?

[brson]

These doc dependencies are also specified here. I think what should be done is that the new -doc-error-index-exec rule should be added to that.

[alexcrichton]

Closing due to inactivity, but feel free to resubmit with comments addressed!