Add use of --package-prefix-hash for testingCommit-type: testingadios: fix pthread dependencyPTScotch: correctly list zlib, pthread as optional dependencies [and not hard-code the libraries her
Add use of --package-prefix-hash for testingCommit-type: testingadios: fix pthread dependencyPTScotch: correctly list zlib, pthread as optional dependencies [and not hard-code the libraries here]configure hash: sort the hash of config files - this way its independent of the order of files listed by 'os.walk'BlasLapack should use self.blaslapack.directory instead of self.libDir since this will be correct with -package-prefix-hashCorrectly handle saving hashvalue to prefix directory when --force is usedPackages that are built after PETSc configure finishes, postProcess() should never go in hash prefix location; just get rebuilt each timeCommit-type: bug-fixReported-by: Satish Balay <balay@mcs.anl.gov>Need to encode hash information before passing to hash library in python3Commit-type: portability-fixReported-by: gitlab CI
show more ...
-package-prefix-hash=directory will prefix the external packages based on a hash of the config directoryand configure argumentsBased on a suggestion by Satish BalayAll PETSc builds go as usual f
-package-prefix-hash=directory will prefix the external packages based on a hash of the config directoryand configure argumentsBased on a suggestion by Satish BalayAll PETSc builds go as usual for testing into PETSC_DIR/PETSC_ARCH.Running with a different branch with the same config will generate the same hash and hencereuse the previous external package builds.Any change in the hash results in a COMPLETELY new external packages prefixIf one assumes that most merge requests do not involve changes to config this will result in a dramaticsavings in configure time plus no longer require someone to manually manage the pre-building of externallibraries and keeping them up to date. The same directory can be used for tests for maint and master sincemaint will generate different hashes and thus have its own copies of the external libraries.Also supports --arch-hash that sets PETSC_ARCH based on the same config hash.Commit-type: testing-fix, featureThanks-to: Satish Balay <balay@mcs.anl.gov>
Remove additional uneeded values from petscconf.h; reorganize handling of default debugger, make it lldb on darwinCommit-type: style-fixThanks-to: Jed Brown <jed@jedbrown.org>
Use manually optimized AVX512 kernels by defaultThis affects only AIJ
Merge remote-tracking branch 'origin/maint'
Merged in balay/slepc-add-gitcommitmaster (pull request #1754)configure: add gitcommitmaster so that petsc/master [or related branches] can default to slepc/masterApproved-by: BarryFSmith <bsmith
Merged in balay/slepc-add-gitcommitmaster (pull request #1754)configure: add gitcommitmaster so that petsc/master [or related branches] can default to slepc/masterApproved-by: BarryFSmith <bsmith@mcs.anl.gov>
fix --download-slepc --prefix=/patch PETSC_ARCH=arch-testReported-by: Pierre Jolivet <pierre.jolivet@enseeiht.fr>fix --with-scalar-type=complex build
configure: add gitcommitmaster so that petsc/master [or related branches] can default to slepc/masterAlso fix self.version string for petsc-dev [use for ex: 3.11.99]Reported-by: Pierre Jolivet vi
configure: add gitcommitmaster so that petsc/master [or related branches] can default to slepc/masterAlso fix self.version string for petsc-dev [use for ex: 3.11.99]Reported-by: Pierre Jolivet via petsc-maint <petsc-maint@mcs.anl.gov>
Allow include quadmath library only if supported by Fortran compiler when Fortran compiler is usedAlso print information in configure summary on availability of __float128Commit-type: bug-fixRep
Allow include quadmath library only if supported by Fortran compiler when Fortran compiler is usedAlso print information in configure summary on availability of __float128Commit-type: bug-fixReported-by: Matthew Knepley <knepley@gmail.com>Thanks-to: Jed Brown <jed@jedbrown.org>
Sys: Include and use limits.h* Remove configure check for C89 limits.h* #define PETSC_BITS_PER_BYTE CHAR_BIT* #define PETSC_IS_COLORING_MAX {UCHAR|USHRT}_MAX
Configure was saving it's dependency information at startup, thus a crash of configure left the dependency informationInstead delete the current dependency file at startup and create a new dependen
Configure was saving it's dependency information at startup, thus a crash of configure left the dependency informationInstead delete the current dependency file at startup and create a new dependency file only after configurehas successfully completedCommit-type: bug-fixReported-by: Lisandro Dalcin <dalcinl@gmail.com>
configure: avoid duplicates, make hash log more readable, and use sha256
configure: Short cut the run if nothing has changed in the config dir, PATH and options.Add '--force' option - to skip this check - and run to completion.sort and remove PETSC_ARCH argument from sy
configure: Short cut the run if nothing has changed in the config dir, PATH and options.Add '--force' option - to skip this check - and run to completion.sort and remove PETSC_ARCH argument from sys.argv that are hashed so works from configure as well as reconfigure scriptexclude '.pyc', #emacs#, .files from hashCommit-type: featureThanks-to: Jed Brown <jed@jedbrown.org>
Minor cleanup of handling of __float128 mostly specifically related to BLAS/LAPACKRemoved if self.skippackagewithoptions: return from Package.consistencyChecks() it is not needed and prevented
Minor cleanup of handling of __float128 mostly specifically related to BLAS/LAPACKRemoved if self.skippackagewithoptions: return from Package.consistencyChecks() it is not needed and prevented fblaslapack.py from performing those checksRemoved redundant checks in fblaslapack.py Install() that are no longer needed since they are now checked with Package.consistencyChecks()Removed redundant check in OpenBLAS.py that is already checked by Package.consistencyChecks()Added self.skippackagewithoptions to OpenBLAS.py since it does not support the --with-pkg optionsAdded better error messages when __float128 was used and the BLAS/LAPACK tests failed. For example in one case it was suggesting using yum or apt (which definitely would not work)Added generation of PETSC_HAVE___FLOAT128 for when __float128 is availableAdded short cut error checking for fblaslapack.py and f2cblaslapack.py when with-64-bit-blas-indices or know-64-bit-blas-indices are used. This prevents the user from having to wait for the download and complete install before knowing.Commit-type: testing-fix, style-fix, feature
configure: When --with-clanguage=cxx is used, verify that the c++ compiler can actually compile .c sources [as c++]Reported-by: Fande Kong <fdkong.jd@gmail.com>
Merge branch 'jed/python3-futurize' [PR #1122]Merge announces support for Python 3.4+ in changes/dev.html.* jed/python3-futurize: (44 commits) gmakegentest: avoid prefix matching in regex dm
Merge branch 'jed/python3-futurize' [PR #1122]Merge announces support for Python 3.4+ in changes/dev.html.* jed/python3-futurize: (44 commits) gmakegentest: avoid prefix matching in regex dm tests ex23.c: fix test args: -one -two -three gmakegentest: more fixes gmakegentest: fix processing of arguments containing number followed by "-" gmakegentest: always loop over nsize; Python-3 fixes python3: add in a python-3.4 test python: fix build on osx/opensolaris - where 'RANLIB=ranlib -c' i.e with a space.. python3: fix --download-ptscotch error configure: glob() gets executed early - whereas we need to get file list later [after 'make all' or equivalent] so revert back to using shell to get file-list python3: map() requires a list modifier for this usage. python: replace ModuleNotFoundError a python3.6+ feature with ImportError config: call setup to ensure that debugLevel is set nightlies: switch linux-pkg-dbg-ftn-interfaces to python3 config: enable Python-3 config: urllib/urlparse updates for Python-3 compatibility config fblaslapack: use isIBM instead of manual test python: replace commands with subprocess Fix LOCDIR paths generateetags.py: switch from commands to subprocess; unicode fix python: don't import commands when it isn't used ...
confgure: fix --with-clean - when --prefix option is usedIt should delete build files - not prefix/installed files.Reported-by: Chris Coutinho <c.coutinho@redstack.nl>
config executeShellCommand: support chained commands and non-shellThe present approach of concatenating strings is problematic if pathscontain spaces or special characters. We should always use l
config executeShellCommand: support chained commands and non-shellThe present approach of concatenating strings is problematic if pathscontain spaces or special characters. We should always use listsinstead of strings, but I don't have the patience to convert all callsites right now. I have converted a representative sample, however,to demonstrate that the approach works. This involved a new functionScript.executeShellCommandSeq() that takes a sequence/list of commandsto be run.A legacy pattern in BuildSystem has been to use chained sequences ofthe form cd to/dir && run commandinstead of using the cwd keyword argument. This commit fixes manysuch instances to use the cwd argument.
config: use open() instead of file() to open filesThis is the recommended usage and file is not defined in Python-3.
config: remove "import user"; deprecated in Python 2.6 and removed in 3I can't imagine why anyone would want this unpredictable behavior in aconfiguration system in the first place.
windows: enable dlls with fortran build
Merge branch 'maint'
windows: disable sharedlibries when fortran compiler is enabled.And if the user explicitly enables sharedlibraries [with fortran compiler] print an appropriate error messageReported-by: Helgad As
windows: disable sharedlibries when fortran compiler is enabled.And if the user explicitly enables sharedlibraries [with fortran compiler] print an appropriate error messageReported-by: Helgad Asdis <asdisprjonar@gmail.com>
dsymutil: check for it only on OSX. And fix check for shared_darwin target (use check for OSX - not dsymutil check)Reported-by: Pedro Marcelino <pedronovaesmelo@gmail.com>Thanks-to: Jed Brown <jed
dsymutil: check for it only on OSX. And fix check for shared_darwin target (use check for OSX - not dsymutil check)Reported-by: Pedro Marcelino <pedronovaesmelo@gmail.com>Thanks-to: Jed Brown <jed@jedbrown.org>
1234567