> Thanks Brandon
> After downloading the source tarball and doing a build successfully I
> wanted to run the testsuite.
You can indeed run the testsuite.
However, note that the testsuite is not included in the "-src" tarball
to keep the distribution size down. There is a separate "-testsuite"
tarball which includes the testsuite/ subtree. This can be extracted
into the parent directory of the source tree. You should find this
tarball in the usual place .
> Are there any guidelines on how much memory is required to run the tests?
> On my Mac I have 12 GB of RAM and a few hundred gigabytes of free disk
> space. The first time I did
> - make THREADS=4 test
> and I got a message
> - Your system has run out of application memory
> so I reran with
> - make test
> and left it running unattended. My machine crashed about an hour later.
Well, unfortunately one of the changes that just barely missesd the
window for -rc1 was a fix to a rather serious memory leak in the compiler
(#13426). I suspect this is the reason you are seeing such high memory
> Ben Gamari wrote:
>> Well, unfortunately one of the changes that just barely missesd the
>> window for -rc1 was a fix to a rather serious memory leak in the compiler
> Thats going to fixed for rc2?
Yes, the patch is already in the ghc-8.2 branch.
tl;dr: If you would like to produce a binary distribution for GHC
8.2.1-rc2 then let me know, grab the source distribution and
start building. The binary distributions will be announced one
week from today.
Hello GHC packagers,
I am happy to announce the release of the 8.2.1-rc2 source distribution
to binary packagers. You will find the usual source artifacts at
As usual, the sooner we can get the binary distributions together the
better, but I will hold off on announcing the distributions until next
Sunday to ensure we're all on the same page. It would be appreciated if
you could reply to this message confirming that you intend to offer a
binary distribution this release.
Otherwise, let me know if you have any trouble building your
distribution. I have yet to push the ghc-8.2.1-rc1 tag in case we
encounter unexpected issues but all of my builds with this tarball
thusfar have gone well.
Note that in addition to the usual complement of Linux/Windows/Darwin
bindists, I have also produced a FreeBSD distribution this time around.
I've noticed that as my tools for producing these distributions improves
the marginal cost of producing another distribution is shrinking. I
would be happy to add OpenBSD as well, but first we'll need to nail
#10032, as far as I understand.
On 05/ 8/17 10:58 PM, Ben Gamari wrote:
> Note that in addition to the usual complement of Linux/Windows/Darwin
> bindists, I have also produced a FreeBSD distribution this time around.
> I've noticed that as my tools for producing these distributions improves
> the marginal cost of producing another distribution is shrinking. I
> would be happy to add OpenBSD as well, but first we'll need to nail
> #10032, as far as I understand.
the situation on OpenBSD is a little bit lucky than on Solaris since I'm
not sure if this helps, but "system" libffi is in fact installed into
/usr/local from ports and there is no other libffi on OpenBSD and this
by probably lucky coincidence makes GHC working with the only quirk
which is GNU tar complaining about missing header files while making
"rm" -f bindistprep/ghc-220.127.116.1170507-x86_64-unknown-openbsd.tar
cd bindistprep && "/usr/local/bin/gtar" hcf - -T ../bindist-list |
/usr/local/bin/xz -c >
/usr/local/bin/gtar: ghc-18.104.22.16870507/rts/dist/build/ffi.h: Cannot
stat: No such file or directory
Cannot stat: No such file or directory
/usr/local/bin/gtar: Exiting with failure status due to previous errors
mv bindistprep/*.tar.xz .
anyway, tarbal is generated, it unpacks well, it even install well (with
just ./configure --prefix=<path>; gmake install) and resulting compiler
is working and linked libffi is where it should be:
$ ghc --make HelloWorld.lhs
[1 of 1] Compiling Main ( HelloWorld.lhs, HelloWorld.o )
Linking HelloWorld ...
In function `copyArg':
warning: warning: strcpy() is almost always misused, please use
/usr/local/lib/libgmp.so.10.0: warning: warning: vsprintf() is often
misused, please use vsnprintf()
In function `showStgWord64':
warning: warning: sprintf() is often misused, please use snprintf()
$ ldd HelloWorld
so if you like you can give it a try on your OpenBSD. If you are using
latest 6.1 please make sure you build and test on wxallowed mount:
/dev/sd1a on / type ffs (local)
/dev/sd1d on /usr/local type ffs (local, nodev, wxallowed)
otherwise you would get strange "No permission error" while executing
any GHC generated executable including tests run by ./configure. I.e.
OpenBSD starts to be more picky about programs which execute code from
writeable memory page and such security sinners need to be run from
wxallowed paths only. Matthias Killian (cced) has done some work on GHC
to fix that but IIRC he hit the wall somewhere in rts (IIRC) so nothing
from this yet.