This is a follow-up to 17192d89 which missed that export.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

...on recently touched files

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

This helps Haddock make tose hyperlinked functions.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

This seemed to be an obvious addition while working on #8433.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

...and while at it, use `bracket` to fix a potential resource leak due
to `freeStablePtr` not being called if `throwErrnoIfMinus1` throws an
exception.

This fixes #8433

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

The removed paragraph is obsolete and does no longer apply to the
implementation as noted by Duncan in #8286.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

On Android, the functions imported from `<termios.h>` are actually
inlined functions, so we need to wrap them via the `capi` calling
convention.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

(And a small matter of using != instead of /= -- how embarrassing!)

On MacOS X, it is defined as "(((__uint64_t)1 << 63) - 1)", and
hence cannot be used inside C preprocessor logic.

CPP conditional code sections have been carefully taken into account
while performing this warning cleanup. Most `OPTIONS_GHC` declarations
could be dropped (now only a lonely `{-# OPTIONS_GHC -fno-cse #-}`
remains in `Posix.Signals`).

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

By avoiding `import Foreign` this accomplishes backward-compatibility
w.r.t. the `unsafePerformIO` relocation w/o resorting to CPP
conditionals.

Moreover, in order to support base<4.7 this commits needs to reintroduce
a harmless backport of `newFilePath`.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

The changelog file will be shown on Hackage once it's included in the
source tarball.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Only language extensions not active when `-XHaskell2010` is enabled are
declared, this way we can drop many redundant
`{-# LANGUAGE ForeignFunctionInterface #-}` occurences. This commit also
removes/replaces some `{-# OPTIONS_GHC -XRecordWildCards #-}` pragmas.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

This is a follow-up to b8fb0a0a which made the module exports of
`tellDirStream` and `seekDirStream` CPP-conditional but didn't
explicitly include `HsUnixConfig.h` which defines the respective CPP
defines.

Signed-off-by: Herbert Valerio Riedel <hvr@gnu.org>

Their use site is now unconditional. Issue #8223.

Signed-off-by: Austin Seipp <austin@well-typed.com>

redneb authored 12 years ago and thoughtpolice committed 11 years ago

Also comes with tests. This closes #8108.

Signed-off-by: Austin Seipp <aseipp@pobox.com>

Submitted by: Nathan Hüsken <nathan.huesken@posteo.de>

Submitted by: Nathan Hüsken <nathan.huesken@posteo.de>

The Bool field of Terminated is new, as is the documentation:

data ProcessStatus
   = Exited ExitCode        -- ^ the process exited by calling
                            -- @exit()@ or returning from @main@
   | Terminated Signal Bool -- ^ the process was terminated by a
                            -- signal, the @Bool@ is @True@ if a core
                            -- dump was produced
   | Stopped Signal         -- ^ the process was stopped by a signal
   deriving (Eq, Ord, Show)

This is an API change, hence will need a major version bump.

Fixes signals004(threaded1,threaded2) on OS X 32.

I assume that this is what was intended.

We were freeing the string, but the string becomes part of the
environment.

and remove code to support older versions

This fixes warnings following
    commit 910a642294eb3547d0cbb3d5735ad81b964f137b
    Author: Simon Peyton Jones <simonpj@microsoft.com>
    Date:   Mon Oct 29 23:25:25 2012 +0000
    Do not treat a constructor in a *pattern* as a *use* of that constructor

I'm not sure if this is the right thing to do: In
    commit 1c4608e3
    Author: Simon Marlow <marlowsd@gmail.com>
    Date:   Thu Feb 19 10:05:32 2009 +0000
    Rewrite of signal-handling.
the commit message says:
    The new implementation has the capability to define signal handlers
    that have access to the siginfo of the signal (#592), but this
    functionality is not exposed in this patch.
but this at least gets validate builds working again, and we can change
it if necessary as part of #2451

redneb authored 13 years ago and pcapriotti committed 13 years ago

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

sleep doesn't really work on GHC because it is always immediately
interrupted by SIGVTALRM used in the RTS.

I explained the problem in a comment and added a WARNING pragma.

usleep and nanosleep have a similar problem, but, since they have better
precision, they can be restarted, so they are not as unusable as sleep.