GitLab

Follows discussion on the libraries@ mailing list; see comments for
rationale.

These files are no longer used

For the moment, this replaces the class (<=) because it is
easier to fit with GHC's coercion evidence.

sys/timeb.h is deprecated on FreeBSD meaning validation fails quite early
without this patch.

This was breaking cross-module inlining, since the symbol ended up being
mangled twice, if the call happened to be inlined into another module.

This reverts commit 7e04ab6158957c90e3c68911b6909b6ef69621da.

Based on a patch by Evan Laforge <qdunkan@gmail.com>

elaforge authored Mar 11, 2012 and pcapriotti committed Apr 09, 2012

Signed-off-by: Paolo Capriotti <p.capriotti@gmail.com>

This avoids the duplication of types and classes for each separate
type of literal: before we had two copies of most functions, one for
numbers and one for strings.  Now we just have one polymorphic (both
at the type and the kind level!) version, which uses a type function
and polymorphic kinds in an interesting way.

This follows a change in GHC and ghc-events. The previous name was misleading.

From commit "Change the presentation of parallel GC work balance in +RTS -s":

    Also rename internal variables to make the names match what they hold.
    The parallel GC work balance is calculated using the total amount of
    memory copied by all GC threads, and the maximum copied by any
    individual thread. You have serial GC when the max is the same as
    copied, and perfectly balanced GC when total/max == n_caps.

Given (T ty1) and ty2, we were computing the fingerprint of the
application (T ty1 ty2) by combining the two fingerprints from (T ty1)
and ty2.  But that gives a different answer to combinging the three
fingerprints from T, ty1, and ty2, which is what happens if you
build the type all at once.  Urk!

Fixes Trac #5962

When there is data in a handle buffer, never fetch more than the
available number of elements, since that can cause a blocking read on
Windows.

See the comment for details.