quasicomputational authored 6 years ago and Alexis Williams committed 6 years ago

These globs make `haddock`, `sdist` and `install` die, so they
definitely ought to be warned about!

Because of the dying behaviour, I made the checks dist-inexcusable;
starting from a clean slate I would probably have only made them
suspicious, but this isn't terrible.

quasicomputational authored 6 years ago and Alexis Williams committed 6 years ago

This also significantly improves the error when trying to refer to
missing directories, hopefully making it clear that it's coming from
Cabal. #5318 and snowleopard/hadrian#634 are two bugs which manifested
as Cabal trying to glob in a non-existent directory and both took some
debugging because of the obscurity of the error.

Previously, we were checking the package with a hard-coded root
directory of ".". This was not a problem before, but with #5372 we
have started to expand globs while checking packages, which breaks if
the CWD is not the directory containing the `.cabal` file and causes
snowleopard/hadrian#634.

Luckily, this is an easy fix: the correct directory is easy to
determine. Writing a test and making sure it's tickling the failing
case took longer than writing the fix!

"." is hard-coded as the root directory passed to `checkPackageFiles`
in a few other places, but those are (a) non-trivial to test, and (b)
already in places that have other assumptions about their CWD, so I
have simply documented the CWD requirement for those.

It's non-obvious that the expected way to pass command line options to
a test suite would be a different command, so mention it explicitly.

Closes #5416.

7.4 is out of our support window, so it's not really important to consider for the status of a PR (and can contribute to error fatigue)

Lexically, they're identical to 2.2, but the semantics of globs have
changed slightly.

quasicomputational authored 6 years ago and Mikhail Glushenkov committed 6 years ago

jgm/zip-archive#47 has been merged and released as part of 0.3.3, so
there is now no need to grab old-time and old-locale in booststrap.sh.

quasicomputational authored 6 years ago and Mikhail Glushenkov committed 6 years ago

This was only a convenience function, but its use could obscure how it
is introducing a dependency on the CWD. By removing it, the "."
argument to `matchDirFileGlob` is explicit.

Any external code using `matchFileGlob` would have needed to be
changed as #5284 changed its signature and the module it lives in; it
is not much more of a burden to switch to `matchDirFileGlob` at the
same time.

quasicomputational authored 6 years ago and Herbert Valerio Riedel committed 6 years ago

As suggested in #5401.

[skip ci]

y[ci skip]

[ci skip]

[ci skip]

Strip empty arguments lists from the elabProgramArgs in
ElaboratedConfiguredPackage to eliminate spurious configuration changes being
found.

Move package flag normalisation code for the new-build hash into it's own
`normaliseConfiguredPackage` function and reuse it to detect whether the
configuration of a local package has changed and needs to be rebuild.

It was being properly wrapped when the command was invoked, but not
for --help output. Hard-wrapping is unfortunate but it's the best we
can do in this scenario; it's better than the previous manual
wrapping, at least.

We don't need to hard-wrap the message as we originate it, since
that's handled elsewhere.

quasicomputational authored 6 years ago and Alexis Williams committed 6 years ago

The code does the right thing, but it's slightly fragile since it
depends on directory switching, so make sure it stays right.