As pointed out in #16085, these ghci tests are fragile in the unregisterised
way.

Alan Zimmerman authored 6 years ago and Ben Gamari committed 6 years ago

For an API annotation to be useful, it must not occur before the span
it is enclosed in.

So, for check-api-annotation output, a line such as

((Test16212.hs:3:22-36,AnnOpenP), [Test16212.hs:3:21]),

should be flagged as an error, as the AnnOpenP location of 3:21
precedes its enclosing span of 3:22-26.

This patch does this.

Closes #16217

It imports System.Posix.IO.

Otherwise the testsuite driver crashes when run multiple times with CLEANUP=NO
on a test containing such extra_files.

Previously `thNameToGhcName` was calling `lookupOrigNameCache` directly, which
failed to handle the case that the name wasn't already in the name cache. This
happens, for instance, when the name was in scope in a plugin being used during
compilation but not in scope in the module being compiled. In this case we the
interface file containing the name won't be loaded and `lookupOrigNameCache`
fails. This was the cause of #16104.

The solution is simple: use the nicely packaged `lookupOrigIO` instead.

Previously testing code-generation for ISA extensions was nearly impossible
since we had no ability to determine whether the host supports the needed
extension. Here we fix this by introducing a simple /proc/cpuinfo-based
testsuite predicate. We really ought to

[skip ci]

[skip ci]

Apparently GitLab CI defaults to declaring all jobs of the previous
stage as dependencies of a job. This meant that we would end up
downloading all of our binary distributions during the `cleanup` stage,
eating up a truly remarkable amount of S3 tranfers.

See #16085.

It's pretty unlikely we will be going back to circleci at this point

[skip-ci]

Alec Theriault authored 6 years ago and Ben Gamari committed 6 years ago

Although the Haddock's for `shiftL` and `shiftR` do require the number
of bits to be non-negative, we should still check this before calling
out to primitives (which also have undefined behaviour for negative bit
shifts).

If a user _really_ wants to bypass checks that the number of bits is
sensible, they already have the aptly-named `unsafeShiftL`/`unsafeShiftR`
at their disposal.

See #16111.

Alec Theriault authored 6 years ago and Ben Gamari committed 6 years ago

Summary:
This adds top-level configure flags '--with-intree-gmp' and
'--with-framework-preferred', both of which are especially relevant
on MacOS.

Besides gaining two new flags, Hadrian also had to be taught what to
do with the 'framework' in .cabal files.

Test Plan:
./boot && ./configure --with-intree-gmp && ./hadrian/build.sh
./boot && ./configure --with-gmp-framework-preferred && ./hadrian/build.sh # on macos

Reviewers: carter, snowleopard, alpmestan, hvr, goldfire, bgamari

Subscribers: rwbarton, erikd

GHC Trac Issues: #16001

Differential Revision: https://phabricator.haskell.org/D5417

See #16218.

This is the last failing test on Darwin preventing us from disallowing CI
failures. See #16201.

Alec Theriault authored 6 years ago and Ben Gamari committed 6 years ago

Since the `$(docdir)` can be picked independently from the `$(libdir)`,
we need to make sure that that the `haddock-html` and `haddock-interface`
fields in the package DB (which is in the `$(libdir)`) get updated to
point to the appropriate places in the `$(docdir)`.

NB: in the make system, `ghc-cabal` would cover this sort of thing by
re-running `configure` on installation, but here we get away with a
couple lines of `sed` and a call to `ghc-pkg recache`.

Fixes #16202.

Ben Gamari authored 6 years ago and Ben Gamari committed 6 years ago

Reviewers: monoidal

Reviewed By: monoidal

Subscribers: monoidal, rwbarton, carter

Differential Revision: https://phabricator.haskell.org/D5056

Joachim Breitner authored 6 years ago and Ben Gamari committed 6 years ago

really minor, but it annoyed me when reading it :-)

Matthew Pickering authored 6 years ago and Ben Gamari committed 6 years ago

In Alp's refactoring of `getProgramContexts` he removed a call to
`getProgramContext` which was where the logic for this used to be
implemented.

Fixes #16214

Simon Peyton Jones authored 6 years ago and Marge Bot committed 6 years ago

The "worker" for a newtype is actually a function
with a small (compulsory) unfolding, namely a cast.

But the construction of this function was plain wrong
for newtype /instances/; it cast the arguemnt to the
family type rather than the representation type.

This never actually bit us because, in the case of a
family instance, we immediately cast the result to
the family type.  So we get
   \x. (x |> co1) |> co2

where the compositio of co1 and co2 is ill-kinded.
However the optimiser (even the simple optimiser)
just collapsed those casts, ignoring the mis-match
in the middle, so we never saw the problem.

Trac #16191 is indeed a dup of #16141; but the resaon
these tickets produce Lint errors is not the unnecessary
forcing; it's because of the ill-typed casts.

This patch fixes the ill-typed casts, properly.  I can't
see a way to trigger an actual failure prior to this
patch, but it's still wrong wrong wrong to have ill-typed
casts, so better to get rid of them.

Samuel Holland authored 6 years ago and Marge Bot committed 6 years ago

GHC has code to handle unsuffixed .so files that are linker scripts
pointing to the real shared library. The detection is done by parsing
the result of `dlerror()` after calling `dlopen()` and looking for
certain error strings. On musl libc, the error message is "Exec format
error", which happens to be `strerror(ENOEXEC)`:

```
$ cat tmp.c
#include <dlfcn.h>
#include <stdio.h>

int main(void) {
        dlopen("libz.so", RTLD_NOW | RTLD_GLOBAL);
        puts(dlerror());
        return 0;
}
$ gcc -o tmp tmp.c
$ ./tmp
Error loading shared library libz.so: Exec format error
$
```

This change fixes the workaround to also work on musl libc.

Link: https://phabricator.haskell.org/D5474

GitLab uses this file to suggest reviewers based upon the files that a Merge
Request touches.

[skip-ci]

This begins to define our expectations of contributions.

[skip-ci]