Bug#1027851: pytorch FTBFS with Python 3.11 as default version
On Mon, 2023-01-30 at 06:46 +0100, Andreas Tille wrote: > Am Sun, Jan 29, 2023 at 10:22:24AM -0500 schrieb M. Zhou: > > > Since we do not have this module[2] (yet) we should probably exclude all > tests that need this module, right? If you think its a nice thing to > have I would volunteer to package this in DPT. Yes, we can skip these python tests for now. And the fix is already uploaded. We should be ready to wait for the testing migration.
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
Am Sun, Jan 29, 2023 at 10:22:24AM -0500 schrieb M. Zhou: > And Aron has uploaded pytorch to NEW. It has cleared new quite quickly but is featuring an autopkgtest regression[1]: Traceback (most recent call last): File "/tmp/autopkgtest-lxc.as10mbia/downtmp/build.o5C/src/test/run_test.py", line 22, in from torch.testing._internal.common_utils import ( File "/usr/lib/python3/dist-packages/torch/testing/_internal/common_utils.py", line 57, in import expecttest ModuleNotFoundError: No module named 'expecttest' Since we do not have this module[2] (yet) we should probably exclude all tests that need this module, right? If you think its a nice thing to have I would volunteer to package this in DPT. Kind regards Andreas. [1] https://ci.debian.net/data/autopkgtest/testing/amd64/p/pytorch/30823657/log.gz [2] https://pypi.org/project/expecttest/ -- http://fam-tille.de
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
On Sun, 2023-01-29 at 09:03 +0100, Andreas Tille wrote: > > > I have no idea about fmtlib but I noticed: > > [2022-09-04] fmtlib 9.1.0+ds1-2 MIGRATED to testing (Debian testing > watch) > [2022-09-04] Accepted fmtlib 9.1.0+ds1-2 (source) into unstable > (Shengjing Zhu) > [2022-08-27] Accepted fmtlib 9.1.0+ds1-1 (source) into experimental > (Shengjing Zhu) > [2022-08-24] fmtlib 9.0.0+ds1-4 MIGRATED to testing (Debian testing > watch) > > Is this failure dating back to August last year and possibly > connected to > the version bump from 9.00 to 9.1.0? I had some similar thoughts during diagnosis. That said, I have already patched the line that FTBFS, and reverted it back to the std::ostringstream implementation, which is merely introducing some overhead. And Aron has uploaded pytorch to NEW.
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
Hi,
Am Sat, Jan 28, 2023 at 03:09:47AM +0800 schrieb Aron Xu:
> On Fri, Jan 27, 2023 at 9:42 PM Andreas Tille wrote:
> >
> > So what help could I (as someone who does not know pytorch at all, just
> > maintains some packages that are depending from it) can I provide?
> >
>
> Here is the fail log just in case you can have a look...
>
> /build/pytorch/build$ ninja
> [1/5] Building CXX object
> caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o
> FAILED: caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o
> /usr/bin/c++ -DAT_PER_OPERATOR_HEADERS -DBUILDING_TESTS
> -DGFLAGS_IS_A_DLL=0 -DGLOG_CUSTOM_PREFIX_SUPPORT
> -DHAVE_MALLOC_USABLE_SIZE=1 -DHAVE_MMAP=1 -DHAVE_SHM_OPEN=1
> -DHAVE_SHM_UNLINK=1 -DMINIZ_DISABLE_ZIP_READER_CRC32_CHECKS
> -DONNXIFI_ENABLE_EXT=1 -DONNX_ML=1 -DONNX_NAMESPACE=onnx
> -DTHP_BUILD_MAIN_LIB -DUSE_C10D -DUSE_C10D_GLOO -DUSE_DISTRIBUTED
> -DUSE_EXTERNAL_MZCRC -DUSE_NUMPY -DUSE_RPC -DUSE_TENSORPIPE
> -DUSE_VALGRIND -D_FILE_OFFSET_BITS=64 -Dtorch_python_EXPORTS
> -I/build/pytorch/build/aten/src -I/build/pytorch/aten/src
> -I/build/pytorch/build -I/build/pytorch
> -I/build/pytorch/cmake/../third_party/benchmark/include
> -I/build/pytorch/debian/foxi -I/build/pytorch/build/debian/foxi
> -I/build/pytorch/torch/.. -I/build/pytorch/torch/../aten/src
> -I/build/pytorch/torch/../aten/src/TH
> -I/build/pytorch/build/caffe2/aten/src
> -I/build/pytorch/build/third_party
> -I/build/pytorch/build/third_party/onnx
> -I/build/pytorch/torch/../third_party/valgrind-headers
> -I/build/pytorch/torch/../third_party/gloo
> -I/build/pytorch/torch/../third_party/onnx
> -I/build/pytorch/torch/../third_party/flatbuffers/include
> -I/build/pytorch/debian/kineto/libkineto/include
> -I/build/pytorch/torch/csrc -I/build/pytorch/torch/csrc/api/include
> -I/build/pytorch/torch/lib -I/build/pytorch/torch/lib/libshm
> -I/build/pytorch/torch/csrc/api -I/build/pytorch/c10/..
> -I/build/pytorch/torch/lib/libshm/../../../torch/lib -isystem
> /build/pytorch/build/third_party/gloo -isystem
> /build/pytorch/cmake/../third_party/gloo -isystem
> /build/pytorch/cmake/../third_party/googletest/googlemock/include
> -isystem /build/pytorch/cmake/../third_party/googletest/googletest/include
> -isystem /usr/include/opencv4 -isystem /usr/include/eigen3 -isystem
> /usr/include/python3.11 -isystem
> /usr/lib/python3/dist-packages/numpy/core/include -Wdate-time
> -D_FORTIFY_SOURCE=2 -g -O2 -ffile-prefix-map=/build/pytorch=.
> -fstack-protector-strong -Wformat -Werror=format-security
> -gsplit-dwarf -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp
> -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE
> -DEDGE_PROFILER_USE_KINETO -O2 -fPIC -Wno-narrowing -Wall -Wextra
> -Werror=return-type -Werror=non-virtual-dtor
> -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds
> -Wno-unknown-pragmas -Wunused-local-typedefs -Wno-unused-parameter
> -Wno-unused-function -Wno-unused-result -Wno-strict-overflow
> -Wno-strict-aliasing -Wno-error=deprecated-declarations
> -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic
> -Wno-error=redundant-decls -Wno-error=old-style-cast
> -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable
> -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math
> -Werror=format -Werror=cast-function-type -Wno-stringop-overflow
> -DHAVE_AVX512_CPU_DEFINITION -DHAVE_AVX2_CPU_DEFINITION -O2 -g
> -DNDEBUG -fPIC -DCAFFE2_USE_GLOO -DTH_HAVE_THREAD -Wno-unused-variable
> -fno-strict-aliasing -Wno-write-strings -Wno-strict-aliasing
> -std=gnu++14 -MD -MT
> caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o -MF
> caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o.d -o
> caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o -c
> /build/pytorch/torch/csrc/Exceptions.cpp
> /build/pytorch/torch/csrc/Exceptions.cpp: In destructor
> 'torch::PyWarningHandler::~PyWarningHandler()':
> /build/pytorch/torch/csrc/Exceptions.cpp:264:23: error: no matching
> function for call to 'format_to(fmt::v9::memory_buffer&,
> torch::PyWarningHandler::~PyWarningHandler()FMT_COMPILE_STRING,
> std::__cxx11::basic_string&, const char*&, uint32_t&)'
> 264 | fmt::format_to(
> | ~~^
> 265 | buf,
> |
> 266 | FMT_STRING("{} (Triggered internally at {}:{}.)"),
> | ~~
> 267 | msg,
> |
> 268 | source_location.file,
> | ~
> 269 | source_location.line);
> | ~
> In file included from /usr/include/fmt/format.h:48,
> from /build/pytorch/torch/csrc/Exceptions.cpp:10:
> /usr/include/fmt/core.h:3233:17: note: candidate: 'template OutputIt, class ... T, typename
> std::enable_if char>::value, int>::type > OutputIt
> fmt::v9::format_to(OutputIt,
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
For reference, a 8 core + 16GB RAM configuration should be able to finish the pytorch compilation timely. The build takes roughly an hour. My observation is based on power9 -- on amd64 it should be something similar. On Sun, 2023-01-29 at 11:09 +0800, Aron Xu wrote: > On Fri, Jan 27, 2023 at 9:42 PM Andreas Tille wrote: > > > > Am Fri, Jan 27, 2023 at 08:21:46PM +0800 schrieb Aron Xu: > > > On Fri, Jan 27, 2023 at 7:12 PM Andreas Tille wrote: > > > > make: *** [debian/rules:83: binary] Terminated > > > > ninja: build stopped: interrupted by user. > > > > > > > > could be a sign for this. Was I to naive to assume Salsa CI could > > > > manage a pytorch build and should we possibly switch this off again? > > > > > > > > > > Not sure but by wild guess it could be caused by running for too long? > > > > I do not think so. Since I was aware that it will take long I have > > adjusted the timeout from 1h (default) to 4h. The log stops a bit after > > 3h. To my experience if timeout is the reason the log ends with this > > information. > > > > Then I guess it could be out-of-memory, the build process is hungry > for RAM and a single cc1plus process can take at least up to 2GB > memory during my quick observation. > > Regards, > Aron >
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
On Fri, Jan 27, 2023 at 9:42 PM Andreas Tille wrote: > > Am Fri, Jan 27, 2023 at 08:21:46PM +0800 schrieb Aron Xu: > > On Fri, Jan 27, 2023 at 7:12 PM Andreas Tille wrote: > > > make: *** [debian/rules:83: binary] Terminated > > > ninja: build stopped: interrupted by user. > > > > > > could be a sign for this. Was I to naive to assume Salsa CI could > > > manage a pytorch build and should we possibly switch this off again? > > > > > > > Not sure but by wild guess it could be caused by running for too long? > > I do not think so. Since I was aware that it will take long I have > adjusted the timeout from 1h (default) to 4h. The log stops a bit after > 3h. To my experience if timeout is the reason the log ends with this > information. > Then I guess it could be out-of-memory, the build process is hungry for RAM and a single cc1plus process can take at least up to 2GB memory during my quick observation. Regards, Aron
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
On Fri, Jan 27, 2023 at 9:42 PM Andreas Tille wrote:
>
> So what help could I (as someone who does not know pytorch at all, just
> maintains some packages that are depending from it) can I provide?
>
Here is the fail log just in case you can have a look...
/build/pytorch/build$ ninja
[1/5] Building CXX object
caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o
FAILED: caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o
/usr/bin/c++ -DAT_PER_OPERATOR_HEADERS -DBUILDING_TESTS
-DGFLAGS_IS_A_DLL=0 -DGLOG_CUSTOM_PREFIX_SUPPORT
-DHAVE_MALLOC_USABLE_SIZE=1 -DHAVE_MMAP=1 -DHAVE_SHM_OPEN=1
-DHAVE_SHM_UNLINK=1 -DMINIZ_DISABLE_ZIP_READER_CRC32_CHECKS
-DONNXIFI_ENABLE_EXT=1 -DONNX_ML=1 -DONNX_NAMESPACE=onnx
-DTHP_BUILD_MAIN_LIB -DUSE_C10D -DUSE_C10D_GLOO -DUSE_DISTRIBUTED
-DUSE_EXTERNAL_MZCRC -DUSE_NUMPY -DUSE_RPC -DUSE_TENSORPIPE
-DUSE_VALGRIND -D_FILE_OFFSET_BITS=64 -Dtorch_python_EXPORTS
-I/build/pytorch/build/aten/src -I/build/pytorch/aten/src
-I/build/pytorch/build -I/build/pytorch
-I/build/pytorch/cmake/../third_party/benchmark/include
-I/build/pytorch/debian/foxi -I/build/pytorch/build/debian/foxi
-I/build/pytorch/torch/.. -I/build/pytorch/torch/../aten/src
-I/build/pytorch/torch/../aten/src/TH
-I/build/pytorch/build/caffe2/aten/src
-I/build/pytorch/build/third_party
-I/build/pytorch/build/third_party/onnx
-I/build/pytorch/torch/../third_party/valgrind-headers
-I/build/pytorch/torch/../third_party/gloo
-I/build/pytorch/torch/../third_party/onnx
-I/build/pytorch/torch/../third_party/flatbuffers/include
-I/build/pytorch/debian/kineto/libkineto/include
-I/build/pytorch/torch/csrc -I/build/pytorch/torch/csrc/api/include
-I/build/pytorch/torch/lib -I/build/pytorch/torch/lib/libshm
-I/build/pytorch/torch/csrc/api -I/build/pytorch/c10/..
-I/build/pytorch/torch/lib/libshm/../../../torch/lib -isystem
/build/pytorch/build/third_party/gloo -isystem
/build/pytorch/cmake/../third_party/gloo -isystem
/build/pytorch/cmake/../third_party/googletest/googlemock/include
-isystem /build/pytorch/cmake/../third_party/googletest/googletest/include
-isystem /usr/include/opencv4 -isystem /usr/include/eigen3 -isystem
/usr/include/python3.11 -isystem
/usr/lib/python3/dist-packages/numpy/core/include -Wdate-time
-D_FORTIFY_SOURCE=2 -g -O2 -ffile-prefix-map=/build/pytorch=.
-fstack-protector-strong -Wformat -Werror=format-security
-gsplit-dwarf -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp
-DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE
-DEDGE_PROFILER_USE_KINETO -O2 -fPIC -Wno-narrowing -Wall -Wextra
-Werror=return-type -Werror=non-virtual-dtor
-Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds
-Wno-unknown-pragmas -Wunused-local-typedefs -Wno-unused-parameter
-Wno-unused-function -Wno-unused-result -Wno-strict-overflow
-Wno-strict-aliasing -Wno-error=deprecated-declarations
-Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic
-Wno-error=redundant-decls -Wno-error=old-style-cast
-fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable
-Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math
-Werror=format -Werror=cast-function-type -Wno-stringop-overflow
-DHAVE_AVX512_CPU_DEFINITION -DHAVE_AVX2_CPU_DEFINITION -O2 -g
-DNDEBUG -fPIC -DCAFFE2_USE_GLOO -DTH_HAVE_THREAD -Wno-unused-variable
-fno-strict-aliasing -Wno-write-strings -Wno-strict-aliasing
-std=gnu++14 -MD -MT
caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o -MF
caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o.d -o
caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o -c
/build/pytorch/torch/csrc/Exceptions.cpp
/build/pytorch/torch/csrc/Exceptions.cpp: In destructor
'torch::PyWarningHandler::~PyWarningHandler()':
/build/pytorch/torch/csrc/Exceptions.cpp:264:23: error: no matching
function for call to 'format_to(fmt::v9::memory_buffer&,
torch::PyWarningHandler::~PyWarningHandler()FMT_COMPILE_STRING,
std::__cxx11::basic_string&, const char*&, uint32_t&)'
264 | fmt::format_to(
| ~~^
265 | buf,
|
266 | FMT_STRING("{} (Triggered internally at {}:{}.)"),
| ~~
267 | msg,
|
268 | source_location.file,
| ~
269 | source_location.line);
| ~
In file included from /usr/include/fmt/format.h:48,
from /build/pytorch/torch/csrc/Exceptions.cpp:10:
/usr/include/fmt/core.h:3233:17: note: candidate: 'template::value, int>::type > OutputIt
fmt::v9::format_to(OutputIt, format_string, T&& ...)'
3233 | FMT_INLINE auto format_to(OutputIt out, format_string
fmt, T&&... args)
| ^
/usr/include/fmt/core.h:3233:17: note: template argument
deduction/substitution failed:
/usr/include/fmt/core.h:3232:11: error: no type named 'type' in
'struct
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
Feel free to break the pytorch reverse dependencies without a transition slot -- we do not need the slot in the current status. The rdeps are already not in testing due to RC bugs and needs some new patchworks. Manual upload is needed for its rebuild. On Fri, 2023-01-27 at 20:19 +0800, Aron Xu wrote: > On Fri, Jan 27, 2023 at 3:48 PM Andreas Tille wrote: > > > > Hi Aron, > > > > Am Fri, Jan 27, 2023 at 02:09:05AM +0800 schrieb Aron Xu: > > > > > > The packaging work of 1.13.1[1] has started on salsa. We still have a > > > failure related to fmtlib before making the package build successfully > > > [5/1781]. Both Mo and I have limited bandwidth here and help is always > > > appreciated. > > > > I've just checked the changelog and noticed: > > > >Bump SOVERSION to 1.13 > > > > but we are in transition freeze. So this needs to be coordinated with > > release team. I guess if we argue that 1.13 will support Python3.11 > > this could be some argument after the decision that this should be the > > supported Python3 version for the next release. > > > > Agreed. And it seems the only rdepends of libtorch1.12 is > python3-torchvision which is owned by us, too, so the update would be > fairly easy to handle. > > Regards, > Aron >
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
Am Fri, Jan 27, 2023 at 08:21:46PM +0800 schrieb Aron Xu: > On Fri, Jan 27, 2023 at 7:12 PM Andreas Tille wrote: > > make: *** [debian/rules:83: binary] Terminated > > ninja: build stopped: interrupted by user. > > > > could be a sign for this. Was I to naive to assume Salsa CI could > > manage a pytorch build and should we possibly switch this off again? > > > > Not sure but by wild guess it could be caused by running for too long? I do not think so. Since I was aware that it will take long I have adjusted the timeout from 1h (default) to 4h. The log stops a bit after 3h. To my experience if timeout is the reason the log ends with this information. > I'm building and testing it with a quite high end configuration that's > able to finish the build stage in a few minutes... Amazing ... So what help could I (as someone who does not know pytorch at all, just maintains some packages that are depending from it) can I provide? Kind regards Andreas. -- http://fam-tille.de
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
On Fri, Jan 27, 2023 at 7:12 PM Andreas Tille wrote: > > BTW, I tried to switch on Salsa-CI which failed[2] and I suspect this > might be due to insufficient memory. At least I've found references in > Web that > > make: *** [debian/rules:83: binary] Terminated > ninja: build stopped: interrupted by user. > > could be a sign for this. Was I to naive to assume Salsa CI could > manage a pytorch build and should we possibly switch this off again? > Not sure but by wild guess it could be caused by running for too long? I'm building and testing it with a quite high end configuration that's able to finish the build stage in a few minutes... Regards, Aron
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
On Fri, Jan 27, 2023 at 3:48 PM Andreas Tille wrote: > > Hi Aron, > > Am Fri, Jan 27, 2023 at 02:09:05AM +0800 schrieb Aron Xu: > > > > The packaging work of 1.13.1[1] has started on salsa. We still have a > > failure related to fmtlib before making the package build successfully > > [5/1781]. Both Mo and I have limited bandwidth here and help is always > > appreciated. > > I've just checked the changelog and noticed: > >Bump SOVERSION to 1.13 > > but we are in transition freeze. So this needs to be coordinated with > release team. I guess if we argue that 1.13 will support Python3.11 > this could be some argument after the decision that this should be the > supported Python3 version for the next release. > Agreed. And it seems the only rdepends of libtorch1.12 is python3-torchvision which is owned by us, too, so the update would be fairly easy to handle. Regards, Aron
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
BTW, I tried to switch on Salsa-CI which failed[2] and I suspect this might be due to insufficient memory. At least I've found references in Web that make: *** [debian/rules:83: binary] Terminated ninja: build stopped: interrupted by user. could be a sign for this. Was I to naive to assume Salsa CI could manage a pytorch build and should we possibly switch this off again? Kind regards Andreas. [2] https://salsa.debian.org/deeplearning-team/pytorch/-/jobs/3859172 Am Fri, Jan 27, 2023 at 08:47:57AM +0100 schrieb Andreas Tille: > Hi Aron, > > Am Fri, Jan 27, 2023 at 02:09:05AM +0800 schrieb Aron Xu: > > > > The packaging work of 1.13.1[1] has started on salsa. We still have a > > failure related to fmtlib before making the package build successfully > > [5/1781]. Both Mo and I have limited bandwidth here and help is always > > appreciated. > > I've just checked the changelog and noticed: > >Bump SOVERSION to 1.13 > > but we are in transition freeze. So this needs to be coordinated with > release team. I guess if we argue that 1.13 will support Python3.11 > this could be some argument after the decision that this should be the > supported Python3 version for the next release. > > Kind regards > Andreas. > > > [1]https://salsa.debian.org/deeplearning-team/pytorch > > -- > http://fam-tille.de > > -- http://fam-tille.de
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
Hi Aron, Am Fri, Jan 27, 2023 at 02:09:05AM +0800 schrieb Aron Xu: > > The packaging work of 1.13.1[1] has started on salsa. We still have a > failure related to fmtlib before making the package build successfully > [5/1781]. Both Mo and I have limited bandwidth here and help is always > appreciated. I've just checked the changelog and noticed: Bump SOVERSION to 1.13 but we are in transition freeze. So this needs to be coordinated with release team. I guess if we argue that 1.13 will support Python3.11 this could be some argument after the decision that this should be the supported Python3 version for the next release. Kind regards Andreas. > [1]https://salsa.debian.org/deeplearning-team/pytorch -- http://fam-tille.de
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
Hi On Fri, Jan 27, 2023 at 1:27 AM Andreas Tille wrote: > > Hi, > > I was checking this bug log and realised, that the "Forwarded" field > links to an old PR[1] that was merged long before this bug was filed. > Checking upstream I realised that the Debian package is lagging behind > upstream releases. > > Could someone please give a status update what might be the plan to get > pytorch back into testing (either be applying the PR patch to the old > version or upgrading to latest upstream which will most probably support > Python 3.11). > The packaging work of 1.13.1[1] has started on salsa. We still have a failure related to fmtlib before making the package build successfully [5/1781]. Both Mo and I have limited bandwidth here and help is always appreciated. [1]https://salsa.debian.org/deeplearning-team/pytorch Regards, Aron
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
Hi, I was checking this bug log and realised, that the "Forwarded" field links to an old PR[1] that was merged long before this bug was filed. Checking upstream I realised that the Debian package is lagging behind upstream releases. Could someone please give a status update what might be the plan to get pytorch back into testing (either be applying the PR patch to the old version or upgrading to latest upstream which will most probably support Python 3.11). I admit I do not feel good to build pytorch on my limited machines neither do I know the potential consequences of a version upgrade. Kind regards Andreas. [1] https://github.com/pytorch/pytorch/pull/81242 -- http://fam-tille.de
Bug#1027851: pytorch FTBFS with Python 3.11 as default version
Source: pytorch Version: 1.12.1-1 Severity: serious Forwarded: https://github.com/pytorch/pytorch/pull/81242 https://buildd.debian.org/status/fetch.php?pkg=pytorch=amd64=1.12.1-1%2Bb2=1672759960=0 ... FAILED: caffe2/torch/CMakeFiles/torch_python.dir/csrc/Stream.cpp.o /usr/bin/c++ -DAT_PER_OPERATOR_HEADERS -DBUILDING_TESTS -DGFLAGS_IS_A_DLL=0 -DGLOG_CUSTOM_PREFIX_SUPPORT -DHAVE_MALLOC_USABLE_SIZE=1 -DHAVE_MMAP=1 -DHAVE_SHM_OPEN=1 -DHAVE_SHM_UNLINK=1 -DMINIZ_DISABLE_ZIP_READER_CRC32_CHECKS -DONNXIFI_ENABLE_EXT=1 -DONNX_ML=1 -DONNX_NAMESPACE=onnx -DTHP_BUILD_MAIN_LIB -DUSE_C10D -DUSE_C10D_GLOO -DUSE_DISTRIBUTED -DUSE_EXTERNAL_MZCRC -DUSE_NUMPY -DUSE_RPC -DUSE_TENSORPIPE -DUSE_VALGRIND -D_FILE_OFFSET_BITS=64 -Dtorch_python_EXPORTS -I/<>/build/aten/src -I/<>/aten/src -I/<>/build -I/<> -I/<>/cmake/../third_party/benchmark/include -I/<>/debian/foxi -I/<>/build/debian/foxi -I/<>/torch/.. -I/<>/torch/../aten/src -I/<>/torch/../aten/src/TH -I/<>/build/caffe2/aten/src -I/<>/build/third_party -I/<>/build/third_party/onnx -I/<>/torch/../third_party/valgrind-headers -I/<>/torch/../third_party/gloo -I/<>/torch/../third_party/onnx -I/<>/torch/../third_party/flatbuffers/include -I/<>/torch/csrc -I/<>/torch/csrc/api/include -I/<>/torch/lib -I/<>/torch/lib/libshm -I/<>/torch/csrc/distributed -I/<>/torch/csrc/api -I/<>/c10/.. -I/<>/torch/lib/libshm/../../../torch/lib -isystem /<>/build/third_party/gloo -isystem /<>/cmake/../third_party/gloo -isystem /<>/cmake/../third_party/googletest/googlemock/include -isystem /<>/cmake/../third_party/googletest/googletest/include -isystem /usr/include/opencv4 -isystem /usr/include/eigen3 -isystem /usr/include/python3.11 -isystem /usr/lib/python3/dist-packages/numpy/core/include -Wdate-time -D_FORTIFY_SOURCE=2 -g -O2 -ffile-prefix-map=/<>=. -fstack-protector-strong -Wformat -Werror=format-security -gsplit-dwarf -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -DEDGE_PROFILER_USE_KINETO -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow -DHAVE_AVX512_CPU_DEFINITION -DHAVE_AVX2_CPU_DEFINITION -O2 -g -DNDEBUG -fPIC -DCAFFE2_USE_GLOO -DTH_HAVE_THREAD -Wno-unused-variable -fno-strict-aliasing -Wno-write-strings -Wno-strict-aliasing -std=gnu++14 -MD -MT caffe2/torch/CMakeFiles/torch_python.dir/csrc/Stream.cpp.o -MF caffe2/torch/CMakeFiles/torch_python.dir/csrc/Stream.cpp.o.d -o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Stream.cpp.o -c /<>/torch/csrc/Stream.cpp In file included from /usr/include/python3.11/Python.h:44, from /usr/include/pybind11/detail/common.h:266, from /usr/include/pybind11/attr.h:13, from /usr/include/pybind11/detail/class.h:12, from /usr/include/pybind11/pybind11.h:13, from /<>/torch/csrc/Stream.cpp:1: /<>/torch/csrc/Stream.cpp: In function ‘void THPStream_init(PyObject*)’: /<>/torch/csrc/Stream.cpp:112:3: error: lvalue required as left operand of assignment 112 | Py_TYPE() = _Type; | ^~~ ...
