[ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[JonY]

[-- Attachment #1: Type: text/plain, Size: 1065 bytes --]

I have just updated the mingw-w64 cross compilers Cygwin 32bit and
64bit to mingw64-*-gcc-5.4.0-2.

Unlike the Cygwin gcc, this version DOES NOT PRESERVER C++ ABI from
gcc4. Please rebuild all your C++ libraries.

==== PLEASE RECOMPILE ALL YOUR C++ CODE FOR ABI COMPATIBILITY ====

The cross compiler can produce Win32 and Win64 native binaries. As a
cross compiler, you may use --host=i686-w64-mingw32 or
--host=x86_64-w64-mingw32 for autotools based source packages to build
for Windows.

              *** CYGWIN-ANNOUNCE UNSUBSCRIBE INFO ***

If you want to unsubscribe from the cygwin-announce mailing list, look
at the "List-Unsubscribe: " tag in the email header of this message.
Send email to the address specified there. It will be in the format:

cygwin-announce-unsubscribe-you=yourdomain.com <at> cygwin.com

If you need more information on unsubscribing, start reading here:

http://sourceware.org/lists.html#unsubscribe-simple

Please read *all* of the information on unsubscribing that is available
starting at this URL.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[Tony Kelman]

> I have just updated the mingw-w64 cross compilers Cygwin 32bit and
> 64bit to mingw64-*-gcc-5.4.0-2.

I'm regretting only testing on 64 bit in my earlier message. I do see a
segfault now when I build Julia with i686-w64-mingw32-gcc 5.4.0-2, that
did not occur with 4.9.2-2. Could 4.9.2-2 be reinstated as prev? And/or
could packaging of GCC 6 be done with a faster turnaround time? Here are
the repro steps for compiling to 32 bit (with apologies for how long it'll
take, it's mostly miscompilation of LLVM causing the problem):

# check build prereqs
curl --version
patch --version
python --version
g++ --version
m4 --version
cmake --version
7z -h
i686-w64-mingw32-gcc --version
i686-w64-mingw32-g++ --version
i686-w64-mingw32-gfortran --version
# end of build prereq checks
git clone https://github.com/JuliaLang/julia -b release-0.5
cd julia
echo 'XC_HOST = i686-w64-mingw32' > Make.user
make -j8 test


At the end of this, roughly an hour later, I see

    JULIA usr/lib/julia/sys.o

Please submit a bug report with steps to reproduce this fault, and any error messages that follow (in their entirety). Thanks.
Exception: EXCEPTION_ACCESS_VIOLATION at 0xc5e821 -- unknown function (ip: 00C5E821)
while loading sysimg.jl, in expression starting on line 6
unknown function (ip: 00C5E820)
unknown function (ip: 00C5ED73)
unknown function (ip: 00C5ED73)
unknown function (ip: 04B86E87)
unknown function (ip: FFFFFFFE)
...


instead of finishing building and executing the tests as expected.

Thanks,
Tony

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[JonY]

[-- Attachment #1.1: Type: text/plain, Size: 816 bytes --]

On 8/16/2016 15:57, Tony Kelman wrote:
>> I have just updated the mingw-w64 cross compilers Cygwin 32bit and
>> 64bit to mingw64-*-gcc-5.4.0-2.
> 
> I'm regretting only testing on 64 bit in my earlier message. I do see a
> segfault now when I build Julia with i686-w64-mingw32-gcc 5.4.0-2, that
> did not occur with 4.9.2-2. Could 4.9.2-2 be reinstated as prev? And/or
> could packaging of GCC 6 be done with a faster turnaround time? Here are
> the repro steps for compiling to 32 bit (with apologies for how long it'll
> take, it's mostly miscompilation of LLVM causing the problem):
> 

Yaakov,

Can you restore 4.9.2-2 as previous? Thanks.

Normally, new versions will first show up for cygwin native gcc before
being handed down to the cross compilers. I'll see about doing a test
version.


[-- Attachment #1.2: 0xD4EBC740.asc --]
[-- Type: application/pgp-keys, Size: 3191 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[Tony Kelman]

> Normally, new versions will first show up for cygwin native gcc before
> being handed down to the cross compilers. I'll see about doing a test
> version.

How long does building the cross-compiler from your cygport take, usually?
I could try that and let you know if just applying the patch from
https://gcc.gnu.org/viewcvs/gcc/trunk/gcc/config/i386/winnt.c?r1=232828&r2=232827&pathrev=232828
would fix matters. Not sure how many other packaging changes you'd need
to go from 5.4.0 to 6.1.0. (And 6.2 should be coming shortly.)

According to the discussion in https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66655
the underlying issue may have also been a problem for cygwin targets,
but I do much less testing there as Julia relies on async IO (via a fork
of libuv) that cygwin doesn't provide.

-Tony

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[JonY]

[-- Attachment #1: Type: text/plain, Size: 707 bytes --]

On 8/18/2016 08:19, Tony Kelman wrote:
>> Normally, new versions will first show up for cygwin native gcc before
>> being handed down to the cross compilers. I'll see about doing a test
>> version.
> 
> How long does building the cross-compiler from your cygport take, usually?
> I could try that and let you know if just applying the patch from
> https://gcc.gnu.org/viewcvs/gcc/trunk/gcc/config/i386/winnt.c?r1=232828&r2=232827&pathrev=232828
> would fix matters. Not sure how many other packaging changes you'd need
> to go from 5.4.0 to 6.1.0. (And 6.2 should be coming shortly.)
> 

Usually a 5-6 hours per target, but my machine is underpowered, testing
would be much appreciated.




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[Tony Kelman]

>> How long does building the cross-compiler from your cygport take, usually?
>> I could try that and let you know if just applying the patch from
>>  https://gcc.gnu.org/viewcvs/gcc/trunk/gcc/config/i386/winnt.c?r1=232828&r2=232827&pathrev=232828
>> would fix matters. Not sure how many other packaging changes you'd need
>> to go from 5.4.0 to 6.1.0. (And 6.2 should be coming shortly.)
> 
>
> Usually a 5-6 hours per target, but my machine is underpowered, testing
> would be much appreciated.

Sadly I seem to get similar-looking access violations even after rebuilding
5.4.0-2 from the -src cygport with the patch for GCC bug 66655 added. So
maybe some other change is also needed to get this working, hopefully one
that is included in GCC 6. I'll have to test a cross-compilation from
opensuse where they already have mingw-w64 GCC 6 packaged and see if things
are in a better shape there.

-Tony

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[Tony Kelman]

> Sadly I seem to get similar-looking access violations even after rebuilding
> 5.4.0-2 from the -src cygport with the patch for GCC bug 66655 added. So
> maybe some other change is also needed to get this working, hopefully one
> that is included in GCC 6. I'll have to test a cross-compilation from
> opensuse where they already have mingw-w64 GCC 6 packaged and see if things
> are in a better shape there.

I also get a segfault when I cross-compile with GCC 6.1 from opensuse 13.2
(needs wine to run this particular bootstrap build step), so this looks
like it's something else. To be sure, could you test the small ipa-ra.tgz
test case from the top of https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66655
with your build? It runs correctly with my build where I've applied the
patch, but the larger Julia+LLVM case doesn't work. So this is something
new, and/or https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68601 wasn't
actually a duplicate of 66655. Will continue investigating.

-Tony


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[JonY]

[-- Attachment #1: Type: text/plain, Size: 1150 bytes --]

On 8/18/2016 20:08, Tony Kelman wrote:
>> Sadly I seem to get similar-looking access violations even after rebuilding
>> 5.4.0-2 from the -src cygport with the patch for GCC bug 66655 added. So
>> maybe some other change is also needed to get this working, hopefully one
>> that is included in GCC 6. I'll have to test a cross-compilation from
>> opensuse where they already have mingw-w64 GCC 6 packaged and see if things
>> are in a better shape there.
> 
> I also get a segfault when I cross-compile with GCC 6.1 from opensuse 13.2
> (needs wine to run this particular bootstrap build step), so this looks
> like it's something else. To be sure, could you test the small ipa-ra.tgz
> test case from the top of https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66655
> with your build? It runs correctly with my build where I've applied the
> patch, but the larger Julia+LLVM case doesn't work. So this is something
> new, and/or https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68601 wasn't
> actually a duplicate of 66655. Will continue investigating.
> 
> -Tony
> 
> 

The -2 build already includes the patch on the 66655 link.



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[Tony Kelman]

> The -2 build already includes the patch on the 66655 link.

Whoops, sorry I missed that, I accidentally downloaded the 5.4.0-1 src
package and build from that.

Another Julia developer also hit this and we're tracking it at
https://github.com/JuliaLang/julia/issues/18123 with a bit more
information.

Now that I got this to happen from an opensuse cross-compile this
looks like it's a GCC problem rather than specific to what version
cygwin has packaged, so I'll wait until we find out more. Will you
see any mingw-labeled bugs on the GCC tracker, or should we
specifically cc you?

> Yaakov,
> 
> Can you restore 4.9.2-2 as previous? Thanks.

In the meantime it would be good to get 4.9.2-2 back as prev if
we can. Thanks!

-Tony

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[Yaakov Selkowitz]

On Thu, 2016-08-18 at 07:13 +0800, JonY wrote:
> Yaakov,
> 
> Can you restore 4.9.2-2 as previous? Thanks.

Done.  Once calm next runs, 4.9.2-2 will be previous, 5.4.0-2 current,
and the previous 5.x test releases have been removed.

--
Yaakov

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[JonY]

[-- Attachment #1: Type: text/plain, Size: 317 bytes --]

On 8/20/2016 05:54, Yaakov Selkowitz wrote:
> On Thu, 2016-08-18 at 07:13 +0800, JonY wrote:
>> Yaakov,
>>
>> Can you restore 4.9.2-2 as previous? Thanks.
> 
> Done.  Once calm next runs, 4.9.2-2 will be previous, 5.4.0-2 current,
> and the previous 5.x test releases have been removed.
> 
>

Thanks!



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[JonY]

[-- Attachment #1: Type: text/plain, Size: 701 bytes --]

On 8/20/2016 02:59, Tony Kelman wrote:
>> The -2 build already includes the patch on the 66655 link.
> 
> Whoops, sorry I missed that, I accidentally downloaded the 5.4.0-1 src
> package and build from that.
> 
> Another Julia developer also hit this and we're tracking it at
> https://github.com/JuliaLang/julia/issues/18123 with a bit more
> information.
> 
> Now that I got this to happen from an opensuse cross-compile this
> looks like it's a GCC problem rather than specific to what version
> cygwin has packaged, so I'll wait until we find out more. Will you
> see any mingw-labeled bugs on the GCC tracker, or should we
> specifically cc you?
> 

Please CC me, thanks.




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[Tony Kelman]

>> Now that I got this to happen from an opensuse cross-compile this
>> looks like it's a GCC problem rather than specific to what version
>> cygwin has packaged, so I'll wait until we find out more. Will you
>> see any mingw-labeled bugs on the GCC tracker, or should we
>> specifically cc you?
>> 
>
> Please CC me, thanks.

GCC's bugzilla couldn't find a match for the gmail reply-to email you
have here, so see https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77333

-Tony

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[JonY]

[-- Attachment #1.1: Type: text/plain, Size: 497 bytes --]

On 02/20/2017 03:32 PM, Tony Kelman wrote:
> My email client is not letting me bottom post this or send it in
> plaintext, so the list might reject it. There's now a patch that
> works at https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77333#c17, are
> you equipped to run the gcc test suite?
> 

Do you have a small test case for this instead of running the entire
test suite? I'm not sure how much time the full test suite takes.

I'll to include the update in the next 5.x release.



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 829 bytes --]

Re: [ANNOUNCEMENT] Updated: mingw64-*-gcc-5.4.0-2 (x86/x86_64)

[Tony Kelman]

> Do you have a small test case for this instead of running the entire
> test suite? I'm not sure how much time the full test suite takes.

Don't have a small test case, but I do have a big one. I've been
verifying that the patch helps in cross-compiles from opensuse
where it's a lot faster to build a patched gcc (and the test case)
than in cygwin.

Large test case (takes about 15 minutes on my year-old laptop):

curl -L http://llvm.org/releases/3.7.1/llvm-3.7.1.src.tar.xz | tar -xJf -
mkdir -p llvmbuild/bin
cd llvmbuild
echo 'set(CMAKE_C_COMPILER gcc)' > NATIVE.cmake
echo 'set(CMAKE_CXX_COMPILER g++)' >> NATIVE.cmake
cp /usr/i686-w64-mingw32/sys-root/mingw/bin/*.dll bin
cmake ../llvm-3.7.1.src -DCMAKE_SYSTEM_NAME=Windows \
  -DCMAKE_BUILD_TYPE=Release -DLLVM_TARGETS_TO_BUILD=host \
  -DCMAKE_C_COMPILER=i686-w64-mingw32-gcc \
  -DCMAKE_CXX_COMPILER=i686-w64-mingw32-g++ \
  -DCROSS_TOOLCHAIN_FLAGS_NATIVE=-DCMAKE_TOOLCHAIN_FILE=$PWD/NATIVE.cmake
make -j`nproc` opt
bin/opt.exe -slp-vectorizer \
  -S ../llvm-3.7.1.src/test/Transforms/SLPVectorizer/X86/vector.ll


That'll give the following sane LLVM IR output on gcc 4.9, or crash on
unpatched 5+.

; ModuleID = '../llvm-3.7.1.src/test/Transforms/SLPVectorizer/X86/vector.ll'
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
target triple = "x86_64-apple-macosx10.8.0"

define void @test(<4 x i32> %in, <4 x i32> %in2) {
  %k = icmp eq <4 x i32> %in, %in2
  ret void
}


> I'll to include the update in the next 5.x release.

Thanks! We're mainly trying to double-check that the patch doesn't
introduce any regressions, cause more problems than it solves.

-Tony

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple