"autogen.pl --force" fails in master tarball due to missing hwloc dirs #7363
Assignees
Labels
Comments
|
autoreconf -f fails the same in hwloc embedded tarball (with additional errors for missing *.in files). |
bgoglin
added a commit
to bgoglin/hwloc
that referenced
this issue
Feb 19, 2020
…d tarball Embedders don't get utils,tests,contrib/*,doc,... directories in their tarball because they build hwloc in embedded mode. However these tarballs cannot be re-autogen'ed unless they contain all directories needed by automake even in non-embedded mode (autogen doesn't know we're configuring hwloc in embedded or standalone mode). OMPI didn't have any issue in the past because they manually copied hwloc embedded directories AND created standalone subdirectories that their automake would dist. Now that they use hwloc as git submodule, bring the proper fix inside hwloc so that OMPI doesn't have to hardwire that list of subdirectories anymore. Refs open-mpi/ompi#7363 Note that pure-hwloc embedded tarballs still cannot be re-autogen'ed because it needs many more standalone files (everything used by configure, basically) but we do not care. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
bgoglin
added a commit
to bgoglin/hwloc
that referenced
this issue
Feb 19, 2020
…dded tarball Build an embedded tarball from the original tarball and use it instead for all tests. The intend is to have our CI detect problems such as open-mpi/ompi#7363 whenever the list of non-embedded dist directories changes. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
bgoglin
added a commit
to bgoglin/hwloc
that referenced
this issue
Feb 19, 2020
…dded tarball Build an embedded tarball from the original tarball and use it instead for all tests. The intend is to have our CI detect problems such as open-mpi/ompi#7363 whenever the list of non-embedded dist directories changes. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr> (cherry picked from commit 9af4e4d361c0f9e0e44a587c4b9b364b724265a6)
bgoglin
added a commit
to bgoglin/hwloc
that referenced
this issue
Feb 19, 2020
…d tarball Embedders don't get utils,tests,contrib/*,doc,... directories in their tarball because they build hwloc in embedded mode. However these tarballs cannot be re-autogen'ed unless they contain all directories needed by automake even in non-embedded mode (autogen doesn't know we're configuring hwloc in embedded or standalone mode). OMPI didn't have any issue in the past because they manually copied hwloc embedded directories AND created standalone subdirectories that their automake would dist. Now that they use hwloc as git submodule, bring the proper fix inside hwloc so that OMPI doesn't have to hardwire that list of subdirectories anymore. Refs open-mpi/ompi#7363 Note that pure-hwloc embedded tarballs still cannot be re-autogen'ed because it needs many more standalone files (everything used by configure, basically) but we do not care. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr> (cherry picked from commit ba6d413)
bgoglin
added a commit
to bgoglin/hwloc
that referenced
this issue
Feb 19, 2020
…d tarball Embedders don't get utils,tests,contrib/*,doc,... directories in their tarball because they build hwloc in embedded mode. However these tarballs cannot be re-autogen'ed unless they contain all directories needed by automake even in non-embedded mode (autogen doesn't know we're configuring hwloc in embedded or standalone mode). OMPI didn't have any issue in the past because they manually copied hwloc embedded directories AND created standalone subdirectories that their automake would dist. Now that they use hwloc as git submodule, bring the proper fix inside hwloc so that OMPI doesn't have to hardwire that list of subdirectories anymore. Refs open-mpi/ompi#7363 Note that pure-hwloc embedded tarballs still cannot be re-autogen'ed because it needs many more standalone files (everything used by configure, basically) but we do not care. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr> (cherry picked from commit ba6d413)
bgoglin
added a commit
to bgoglin/hwloc
that referenced
this issue
Feb 19, 2020
…dded tarball Build an embedded tarball from the original tarball and use it instead for all tests. The intend is to have our CI detect problems such as open-mpi/ompi#7363 whenever the list of non-embedded dist directories changes. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr> (cherry picked from commit 9af4e4d361c0f9e0e44a587c4b9b364b724265a6)
|
This should be fixed in hwloc commit bgoglin/hwloc@18caedf (from hwloc v2.1) |
bgoglin
added a commit
to bgoglin/hwloc
that referenced
this issue
Feb 20, 2020
…d tarball Embedders don't get utils,tests,contrib/*,doc,... directories in their tarball because they build hwloc in embedded mode. However these tarballs cannot be re-autogen'ed unless they contain all directories needed by automake even in non-embedded mode (autogen doesn't know we're configuring hwloc in embedded or standalone mode). OMPI didn't have any issue in the past because they manually copied hwloc embedded directories AND created standalone subdirectories that their automake would dist. Now that they use hwloc as git submodule, bring the proper fix inside hwloc so that OMPI doesn't have to hardwire that list of subdirectories anymore. Refs open-mpi/ompi#7363 Note that pure-hwloc embedded tarballs still cannot be re-autogen'ed because it needs many more standalone files (everything used by configure, basically) but we do not care. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
bgoglin
added a commit
to bgoglin/hwloc
that referenced
this issue
Feb 20, 2020
…dded tarball Build an embedded tarball from the original tarball and use it instead for all tests. The intend is to have our CI detect problems such as open-mpi/ompi#7363 whenever the list of non-embedded dist directories changes. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
|
@bgoglin I confirm that bgoglin/hwloc@18caedf fixes the issue. Would it be possible to get this merged to the hwloc v2.1 branch? |
|
I'll decide after seeing your votes this morning :) |
|
OK, zapproved. |
bgoglin
added a commit
to open-mpi/hwloc
that referenced
this issue
Feb 21, 2020
…ded tarball Embedders don't get utils,tests,contrib/*,doc,... directories in their tarball because they build hwloc in embedded mode. However it means their tarballs cannot be re-autogen'ed because automake fails when those directories are missing (automake doesn't know if we're in embedded or standalone mode). OMPI didn't have any issue in the past because they manually copied hwloc embedded directories AND created standalone subdirectories that their automake would dist. Now that they use hwloc as git submodule, bring the proper fix inside hwloc so that OMPI doesn't have to hardwire that list of subdirectories anymore. Refs open-mpi/ompi#7363 Note that pure-hwloc embedded tarballs still cannot be re-autogen'ed because they needs many more standalone files (e.g. all .in files used by configure) but we do not care. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
bgoglin
added a commit
to open-mpi/hwloc
that referenced
this issue
Feb 21, 2020
…dded tarball Build an embedded tarball from the original tarball and use it instead for all tests. The intend is to have our CI detect problems such as open-mpi/ompi#7363 whenever the list of non-embedded dist directories changes. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr>
bgoglin
added a commit
to open-mpi/hwloc
that referenced
this issue
Feb 21, 2020
…ded tarball Embedders don't get utils,tests,contrib/*,doc,... directories in their tarball because they build hwloc in embedded mode. However it means their tarballs cannot be re-autogen'ed because automake fails when those directories are missing (automake doesn't know if we're in embedded or standalone mode). OMPI didn't have any issue in the past because they manually copied hwloc embedded directories AND created standalone subdirectories that their automake would dist. Now that they use hwloc as git submodule, bring the proper fix inside hwloc so that OMPI doesn't have to hardwire that list of subdirectories anymore. Refs open-mpi/ompi#7363 Note that pure-hwloc embedded tarballs still cannot be re-autogen'ed because they needs many more standalone files (e.g. all .in files used by configure) but we do not care. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr> (cherry picked from commit b332c93)
bgoglin
added a commit
to open-mpi/hwloc
that referenced
this issue
Feb 21, 2020
…dded tarball Build an embedded tarball from the original tarball and use it instead for all tests. The intend is to have our CI detect problems such as open-mpi/ompi#7363 whenever the list of non-embedded dist directories changes. Signed-off-by: Brice Goglin <Brice.Goglin@inria.fr> (cherry picked from commit 88fc5b4)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
If you run
./autogen.pl --forcein a distribution tarball, it fails because there are hwloc directories not present:We used to include these directories -- along with dummy
Makefile.amfiles -- before we made hwloc be a git submodule.make distcheckworks on a master git checkout, but something is not translating correctly to the generated tarball such that if you run autogen in it, 💥.The text was updated successfully, but these errors were encountered: