LLVM 3.2 failure

Austin Seipp aseipp at pobox.com
Thu Mar 14 17:54:13 CET 2013


The LLVM 3.2 tarball has an annoying bug: it specifies the version as
'3.2svn' and not 3.2. So it's kind of difficult to distinguish them.
You can verify this by downloading the 3.2 tarball from their website
and looking at autoconf's AC_INIT line:

$ pwd
/Users/a/Downloads/llvm-3.2.src
$ grep 3.2svn autoconf/configure.ac
AC_INIT([LLVM],[3.2svn],[http://llvm.org/bugs/])

It's likely Jan is using the right version. It's annoying as hell this
bug is there, though* and LLVM developers don't generally do
point-releases or update the tarballs. It's probably stuck like this
until LLVM 3.3.

* Any interested parties can find a patch for the 3.2 tarball here,
but you'll of course have to apply manually and rebuild:
https://github.com/thoughtpolice/homebrew/blob/35d39a504e619a3443abae0e249b366cc0ae4428/Library/Formula/llvm.rb#L108


On Thu, Mar 14, 2013 at 11:47 AM, Geoffrey Mainland
<mainland at apeiron.net> wrote:
> On 03/14/2013 04:40 PM, Jan Stolarek wrote:
>>> If you type llc -version at the command line, it really says it's 3.2?
>> You don't seem to believe me :)
>
> Given that Austin and I have the stage 2 compiler failure and you don't,
> I think it is reasonable do double check :)
>
>> [killy at xerxes : ~] llc --version
>> LLVM (http://llvm.org/):
>>   LLVM version 3.2svn
>>   Optimized build with assertions.
>>   Built Mar 14 2013 (09:02:06).
>>   Default target: x86_64-unknown-linux-gnu
>>   Host CPU: corei7
>>
>>   Registered Targets:
>>     arm      - ARM
>>     cellspu  - STI CBEA Cell SPU [experimental]
>>     cpp      - C++ backend
>>     hexagon  - Hexagon
>>     mblaze   - MBlaze
>>     mips     - Mips
>>     mips64   - Mips64 [experimental]
>>     mips64el - Mips64el [experimental]
>>     mipsel   - Mipsel
>>     msp430   - MSP430 [experimental]
>>     nvptx    - NVIDIA PTX 32-bit
>>     nvptx64  - NVIDIA PTX 64-bit
>>     ppc32    - PowerPC 32
>>     ppc64    - PowerPC 64
>>     sparc    - Sparc
>>     sparcv9  - Sparc V9
>>     thumb    - Thumb
>>     x86      - 32-bit X86: Pentium-Pro and above
>>     x86-64   - 64-bit X86: EM64T and AMD64
>>     xcore    - XCore
>> [killy at xerxes : ~] opt --version
>> LLVM (http://llvm.org/):
>>   LLVM version 3.2svn
>>   Optimized build with assertions.
>>   Built Mar 14 2013 (09:02:06).
>>   Default target: x86_64-unknown-linux-gnu
>>   Host CPU: corei7
>>
>> So at this point we are clearly dealing with a system-specific
> problem. The possible differences
>> that come to my mind are:
>> - I'm using LLVM 3.2 compiled from source, while you might be using a
> pre-built version from the
>> repository
>> - And I'm also using GHC 7.6.2 that I compiled by myself, instead of
> pre-built binaries available
>> at GHC web site. Are you using the binaries or do you also compiled
> your GHC from sources?
>>
>> Janek
>
> I built LLVM 3.2 from source, but from the release tarball, not
> subversion. Does your svn checkout correspond exactly to the source in
> the 3.2 release tarball?
>
> I also built both GHC 7.4.2 and 7.6.2 from source (release tarballs),
> both using the native back end. Since it's the stage 2 compiler that is
> failing, it's difficult to see why this would matter.
>
> Geoff
>



--
Regards,
Austin



More information about the ghc-devs mailing list