Bug#1024795: python-llvmlite

[M. Zhou]

Control: tags -1 +moreinfo

I'm confused. How did you manage to build the package from source
using c65b3e662b7b08920172b710419d7a06b660be59 on salsa?

I did not upload it because I can never successfully build it
from source.

-------------------------------------------------------------

passmanagers.cpp: In function ‘void LLVMPY_SetTimePasses(bool)’:
passmanagers.cpp:36:37: error: ‘TimePassesIsEnabled’ was not declared in this scope
   36 | LLVMPY_SetTimePasses(bool enable) { TimePassesIsEnabled = enable; }
      |                                     ^~~~~~~~~~~~~~~~~~~
passmanagers.cpp: In function ‘void LLVMPY_AddDeadCodeEliminationPass(LLVMPassManagerRef)’:
passmanagers.cpp:158:50: error: cannot convert ‘llvm::FunctionPass*’ to ‘llvm::Pass*’
  158 |     unwrap(PM)->add(createDeadCodeEliminationPass());
      |                     ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~
      |                                                  |
      |                                                  llvm::FunctionPass*
In file included from passmanagers.cpp:10:
/usr/lib/llvm-14/include/llvm/IR/LegacyPassManager.h:48:26: note:   initializing argument 1 of ‘virtual void
llvm::legacy::PassManagerBase::add(llvm::Pass*)’
   48 |   virtual void add(Pass *P) = 0;
      |                    ~~~~~~^
passmanagers.cpp: In function ‘void LLVMPY_AddSROAPass(LLVMPassManagerRef)’:
passmanagers.cpp:181:75: error: cannot convert ‘llvm::FunctionPass*’ to ‘llvm::Pass*’
  181 | LLVMPY_AddSROAPass(LLVMPassManagerRef PM) { unwrap(PM)->add(createSROAPass()); }
      |                                                             ~~~~~~~~~~~~~~^~
      |                                                                           |
      |                                                                           llvm::FunctionPass*
/usr/lib/llvm-14/include/llvm/IR/LegacyPassManager.h:48:26: note:   initializing argument 1 of ‘virtual void
llvm::legacy::PassManagerBase::add(llvm::Pass*)’
   48 |   virtual void add(Pass *P) = 0;
      |                    ~~~~~~^
make[1]: *** [Makefile.linux:22: passmanagers.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make[1]: Leaving directory '/<<PKGBUILDDIR>>/ffi'
14.0.6

SVML not detected
Traceback (most recent call last):
  File "/<<PKGBUILDDIR>>/ffi/build.py", line 227, in <module>
    main()
  File "/<<PKGBUILDDIR>>/ffi/build.py", line 217, in main
    main_posix('linux', '.so')
  File "/<<PKGBUILDDIR>>/ffi/build.py", line 209, in main_posix
    subprocess.check_call(['make', '-f', makefile] + makeopts)
  File "/usr/lib/python3.10/subprocess.py", line 369, in check_call
    raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['make', '-f', 'Makefile.linux', '-j8']' returned non-zero exit status 2.
error: command '/usr/bin/python3' failed with exit code 1
E: pybuild pybuild:386: build: plugin distutils failed with: exit code=1: /usr/bin/python3 setup.py build 
dh_auto_build: error: pybuild --build -i python{version} -p "3.11 3.10" returned exit code 13
make: *** [debian/rules:11: build] Error 25
dpkg-buildpackage: error: debian/rules build subprocess returned exit status 2


On Thu, 2022-12-22 at 18:20 -0500, M. Zhou wrote:
> Sure, I think we can ship a snapshot version as long as it works
> fine with llvm-14. Could you please verify the snapshot hash
> again?
> 
> https://github.com/numba/llvmlite/commit/c65b3e662b7b08920172b710419d7a06b660be59
> 
> The commit seems missing. If it was close to the master branch,
> I can directly pull the master branch and upload the corresponding
> snapshot.
> 
> On Wed, 2022-12-21 at 20:05 -0800, Diane Trout wrote:
> > Hi,
> > 
> > I was trying to update numba, and need the updated version of llvmlite
> > built against llvm-14
> > 
> > The version that's currently in unstable is still built against llvml-
> > 11
> > 
> > https://packages.debian.org/sid/python3-llvmlite
> > 
> > I've checked out out the llvmlite repository and rebuilt it locally
> > using commit c65b3e662b7b08920172b710419d7a06b660be59 against llvm-14
> > 
> > and llvmlite's test cases pass. (And most of numba's passed too. And I
> > think the remaining test failures aren't related to llvmlite)
> > 
> > Is there a chance we could get an updated version released soon?
> > 
> > Thanks
> > Diane Trout
>