Re: NEW: devel/re2
On 2021/09/08 23:09, Omar Polo wrote:
>
> Stuart Henderson writes:
>
> > On 2021/09/08 22:07, Omar Polo wrote:
> >>
> >> Denis Fondras writes:
> >>
> >> > RE2 is a fast, safe, thread-friendly C++ library alternative to
> >> > backtracking regular expression engines like those used in PCRE,
> >> > Perl, and Python.
> >> >
> >> > https://github.com/google/re2
> >> >
> >> > [2. application/x-tar-gz; re2-20210901.tgz]...
> >>
> >> Hello,
> >>
> >> I had a local port for an earlier version of re2 because it was a
> >> dependency of something I was working on (that rotted on mystuff
> >> probably)
> >>
> >> Anyway, the ports needs some adjustments:
> >>
> >> - make port-lib-depends-check complains about missing WANTLIB on
> >>``${COMPILER_LIBCXX} m''
> >> - since it's a c++ library it's probably more useful to install also
> >>the re2Config.cmake file probably
> >> - it installs the library as lib/libre2.so.9 and libre2.so.9.0.0: in
> >>ports we usually override upstream version with a sensible one
> >>(starting at 0 for new ports)
> >> - it's missing the COMPILER line: it needs C++11 so `base-clang
> >>ports-gcc' should be enough
> >
> > all correct
> >
> >> - the patch for the makefile isn't needed, we can override CXX with
> >>
> >>MAKE_ENV += CXX=${CXX}
> >
> > It uses = in the Makefile so it needs MAKE_FLAGS rather than MAKE_ENV
> > (flags overrides variables set in the makefile, but isn't typically
> > propagated to submakes. env does not override but is propagated. so
> > both are needed sometimes, but usually flags is best).
> >
> >> Attaching a tarball with the mentioned stuff fixed and the build
> >> switched from the provided makefile to cmake (that installs the
> >> re2Config.cmake file and picks the correct C++ compiler).
> >>
> >> (The CMakefile needed some massaging to avoid hardcoding the soname)
> >
> > As-is the version number is not set in the filename of the produced
> > library. This is because the name in SHARED_LIBS has a "lib" prefix which
> > shouldn't be there. Remove the patch and set the correct SHARED_LIBS
> > then the cmake module will take care of it.
>
> that's correct, I didn't noticed the lib prefix, apologies. Attaching
> an updated tarball.
This one is OK sthen@ to import.
> P.S.: I see that there was already a devel/re2 port in openbsd-wip by
> bcallah@
>
ah the write-only repo ;) I prefer the DISTNAME construction in your version
fwiw.
Re: NEW: devel/re2
Stuart Henderson writes:
> On 2021/09/08 22:07, Omar Polo wrote:
>>
>> Denis Fondras writes:
>>
>> > RE2 is a fast, safe, thread-friendly C++ library alternative to
>> > backtracking regular expression engines like those used in PCRE,
>> > Perl, and Python.
>> >
>> > https://github.com/google/re2
>> >
>> > [2. application/x-tar-gz; re2-20210901.tgz]...
>>
>> Hello,
>>
>> I had a local port for an earlier version of re2 because it was a
>> dependency of something I was working on (that rotted on mystuff
>> probably)
>>
>> Anyway, the ports needs some adjustments:
>>
>> - make port-lib-depends-check complains about missing WANTLIB on
>>``${COMPILER_LIBCXX} m''
>> - since it's a c++ library it's probably more useful to install also
>>the re2Config.cmake file probably
>> - it installs the library as lib/libre2.so.9 and libre2.so.9.0.0: in
>>ports we usually override upstream version with a sensible one
>>(starting at 0 for new ports)
>> - it's missing the COMPILER line: it needs C++11 so `base-clang
>>ports-gcc' should be enough
>
> all correct
>
>> - the patch for the makefile isn't needed, we can override CXX with
>>
>> MAKE_ENV += CXX=${CXX}
>
> It uses = in the Makefile so it needs MAKE_FLAGS rather than MAKE_ENV
> (flags overrides variables set in the makefile, but isn't typically
> propagated to submakes. env does not override but is propagated. so
> both are needed sometimes, but usually flags is best).
>
>> Attaching a tarball with the mentioned stuff fixed and the build
>> switched from the provided makefile to cmake (that installs the
>> re2Config.cmake file and picks the correct C++ compiler).
>>
>> (The CMakefile needed some massaging to avoid hardcoding the soname)
>
> As-is the version number is not set in the filename of the produced
> library. This is because the name in SHARED_LIBS has a "lib" prefix which
> shouldn't be there. Remove the patch and set the correct SHARED_LIBS
> then the cmake module will take care of it.
that's correct, I didn't noticed the lib prefix, apologies. Attaching
an updated tarball.
P.S.: I see that there was already a devel/re2 port in openbsd-wip by
bcallah@
re2.tar.gz
Description: Binary data
Re: NEW: devel/re2
On 2021/09/08 22:07, Omar Polo wrote:
>
> Denis Fondras writes:
>
> > RE2 is a fast, safe, thread-friendly C++ library alternative to
> > backtracking regular expression engines like those used in PCRE,
> > Perl, and Python.
> >
> > https://github.com/google/re2
> >
> > [2. application/x-tar-gz; re2-20210901.tgz]...
>
> Hello,
>
> I had a local port for an earlier version of re2 because it was a
> dependency of something I was working on (that rotted on mystuff
> probably)
>
> Anyway, the ports needs some adjustments:
>
> - make port-lib-depends-check complains about missing WANTLIB on
>``${COMPILER_LIBCXX} m''
> - since it's a c++ library it's probably more useful to install also
>the re2Config.cmake file probably
> - it installs the library as lib/libre2.so.9 and libre2.so.9.0.0: in
>ports we usually override upstream version with a sensible one
>(starting at 0 for new ports)
> - it's missing the COMPILER line: it needs C++11 so `base-clang
>ports-gcc' should be enough
all correct
> - the patch for the makefile isn't needed, we can override CXX with
>
> MAKE_ENV += CXX=${CXX}
It uses = in the Makefile so it needs MAKE_FLAGS rather than MAKE_ENV
(flags overrides variables set in the makefile, but isn't typically
propagated to submakes. env does not override but is propagated. so
both are needed sometimes, but usually flags is best).
> Attaching a tarball with the mentioned stuff fixed and the build
> switched from the provided makefile to cmake (that installs the
> re2Config.cmake file and picks the correct C++ compiler).
>
> (The CMakefile needed some massaging to avoid hardcoding the soname)
As-is the version number is not set in the filename of the produced
library. This is because the name in SHARED_LIBS has a "lib" prefix which
shouldn't be there. Remove the patch and set the correct SHARED_LIBS
then the cmake module will take care of it.
Re: NEW: devel/re2
Denis Fondras writes:
> RE2 is a fast, safe, thread-friendly C++ library alternative to
> backtracking regular expression engines like those used in PCRE,
> Perl, and Python.
>
> https://github.com/google/re2
>
> [2. application/x-tar-gz; re2-20210901.tgz]...
Hello,
I had a local port for an earlier version of re2 because it was a
dependency of something I was working on (that rotted on mystuff
probably)
Anyway, the ports needs some adjustments:
- make port-lib-depends-check complains about missing WANTLIB on
``${COMPILER_LIBCXX} m''
- since it's a c++ library it's probably more useful to install also
the re2Config.cmake file probably
- it installs the library as lib/libre2.so.9 and libre2.so.9.0.0: in
ports we usually override upstream version with a sensible one
(starting at 0 for new ports)
- it's missing the COMPILER line: it needs C++11 so `base-clang
ports-gcc' should be enough
- the patch for the makefile isn't needed, we can override CXX with
MAKE_ENV += CXX=${CXX}
Attaching a tarball with the mentioned stuff fixed and the build
switched from the provided makefile to cmake (that installs the
re2Config.cmake file and picks the correct C++ compiler).
(The CMakefile needed some massaging to avoid hardcoding the soname)
Cheers,
Omar Polo
re2.tar.gz
Description: Binary data
NEW: devel/re2
RE2 is a fast, safe, thread-friendly C++ library alternative to backtracking regular expression engines like those used in PCRE, Perl, and Python. https://github.com/google/re2 re2-20210901.tgz Description: application/tar-gz
Ping Again - Re: [New] devel/re2
Ping again, hoping someone can import this one. Thanks. re2-2021.02.02.tgz Description: Binary data Stuart Henderson writes: > this is OK with me to import. > > > On 2021/02/28 15:52, Ashton Fagg wrote: >> Ping. >> > > >> >> >> Ashton Fagg writes: >> >> > It appears I forgot about this, :-). >> > >> > Latest tarball attached - updated to latest version. >> > >> > Thanks. >> > >> > >> > >> > Ashton Fagg writes: >> > >> >> Ping on this one. Tarball attached. >> >> >> >> Ashton Fagg writes: >> >> >> >>> Stuart Henderson writes: >> >>> >> You mentioned a cmakefile change that would increase it, can you show >> that please, some of us only have slow machines like >> >> cpu3: Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz, 3392.17 MHz, 06-3c-03 >> >> :) >> >>> >> >>> Here you go. I've added a patch that increases the timeout, with >> >>> reference in the Makefile. I've increased it from 1500 seconds to >> >>> 2000. This seems to help on my slower machine.
Re: Ping - Re: [New] devel/re2
this is OK with me to import. On 2021/02/28 15:52, Ashton Fagg wrote: > Ping. > > > > Ashton Fagg writes: > > > It appears I forgot about this, :-). > > > > Latest tarball attached - updated to latest version. > > > > Thanks. > > > > > > > > Ashton Fagg writes: > > > >> Ping on this one. Tarball attached. > >> > >> Ashton Fagg writes: > >> > >>> Stuart Henderson writes: > >>> > You mentioned a cmakefile change that would increase it, can you show > that please, some of us only have slow machines like > > cpu3: Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz, 3392.17 MHz, 06-3c-03 > > :) > >>> > >>> Here you go. I've added a patch that increases the timeout, with > >>> reference in the Makefile. I've increased it from 1500 seconds to > >>> 2000. This seems to help on my slower machine.
Ping - Re: [New] devel/re2
Ping. re2-2021.02.02.tgz Description: Binary data Ashton Fagg writes: > It appears I forgot about this, :-). > > Latest tarball attached - updated to latest version. > > Thanks. > > > > Ashton Fagg writes: > >> Ping on this one. Tarball attached. >> >> Ashton Fagg writes: >> >>> Stuart Henderson writes: >>> You mentioned a cmakefile change that would increase it, can you show that please, some of us only have slow machines like cpu3: Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz, 3392.17 MHz, 06-3c-03 :) >>> >>> Here you go. I've added a patch that increases the timeout, with >>> reference in the Makefile. I've increased it from 1500 seconds to >>> 2000. This seems to help on my slower machine.
Re: Ping - Re: [New] devel/re2
It appears I forgot about this, :-). Latest tarball attached - updated to latest version. Thanks. re2.tgz Description: Binary data Ashton Fagg writes: > Ping on this one. Tarball attached. > > Ashton Fagg writes: > >> Stuart Henderson writes: >> >>> You mentioned a cmakefile change that would increase it, can you show >>> that please, some of us only have slow machines like >>> >>> cpu3: Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz, 3392.17 MHz, 06-3c-03 >>> >>> :) >> >> Here you go. I've added a patch that increases the timeout, with >> reference in the Makefile. I've increased it from 1500 seconds to >> 2000. This seems to help on my slower machine.
Ping - Re: [New] devel/re2
Ping on this one. Tarball attached. Ashton Fagg writes: > Stuart Henderson writes: > >> You mentioned a cmakefile change that would increase it, can you show >> that please, some of us only have slow machines like >> >> cpu3: Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz, 3392.17 MHz, 06-3c-03 >> >> :) > > Here you go. I've added a patch that increases the timeout, with > reference in the Makefile. I've increased it from 1500 seconds to > 2000. This seems to help on my slower machine. re2.tgz Description: Binary data
Re: [New] devel/re2
Stuart Henderson writes: > You mentioned a cmakefile change that would increase it, can you show > that please, some of us only have slow machines like > > cpu3: Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz, 3392.17 MHz, 06-3c-03 > > :) Here you go. I've added a patch that increases the timeout, with reference in the Makefile. I've increased it from 1500 seconds to 2000. This seems to help on my slower machine. re2.tgz Description: Binary data
Re: [New] devel/re2
On 2020/11/30 20:05, Ashton Fagg wrote: > Stuart Henderson writes: > > > Could you add a patch to set that to what's needed on your machine please, > > and a comment in the Makefile pointing at it? > > So just yesterday I finally got around to migrating my dev box to better > hardware, and on that machine it now passes with room to spare with the > default time limit. I've added a comment to the Makefile noting why > there may be a timeout (tarball attached). > You mentioned a cmakefile change that would increase it, can you show that please, some of us only have slow machines like cpu3: Intel(R) Xeon(R) CPU E3-1225 v3 @ 3.20GHz, 3392.17 MHz, 06-3c-03 :)
Re: [New] devel/re2
Stuart Henderson writes: > Could you add a patch to set that to what's needed on your machine please, > and a comment in the Makefile pointing at it? So just yesterday I finally got around to migrating my dev box to better hardware, and on that machine it now passes with room to spare with the default time limit. I've added a comment to the Makefile noting why there may be a timeout (tarball attached). re2.tgz Description: Binary data
Re: [New] devel/re2
Stuart Henderson writes: > Could you add a patch to set that to what's needed on your machine > please, and a comment in the Makefile pointing at it? Certainly. I'll get back to you sometime in the next day or two once I have some more time.
Re: [New] devel/re2
On 2020/11/30 10:20, Ashton Fagg wrote: > Stuart Henderson writes: > > > Generally looks good, I do see a timeout on one of the test though, > > do you get the same? > > Yep, same problem here with the default setting. There is a setting that > can be adjusted in the CMakeLists to make it so that the timeout is > longer, but it's hard to say exactly one should set that to. > > I mentioned that in the original mail I sent that I didn't want to make > assumptions about it because I wasn't sure what the best trade-off was, > so any guidance would be appreciated. > > Ash > Could you add a patch to set that to what's needed on your machine please, and a comment in the Makefile pointing at it?
Re: [New] devel/re2
Stuart Henderson writes: > Generally looks good, I do see a timeout on one of the test though, > do you get the same? Yep, same problem here with the default setting. There is a setting that can be adjusted in the CMakeLists to make it so that the timeout is longer, but it's hard to say exactly one should set that to. I mentioned that in the original mail I sent that I didn't want to make assumptions about it because I wasn't sure what the best trade-off was, so any guidance would be appreciated. Ash
Re: [New] devel/re2
On 2020/11/29 15:23, Ashton Fagg wrote: > Fixed. > > Update tarball attached. Sorry for the slow response. Generally looks good, I do see a timeout on one of the test though, do you get the same? ===> Regression tests for re2-20201101 [0/1] cd /usr/obj/ports/re2-20201101/build-amd64 && /usr/local/bin/ctest --force-new-ctest-process --exclude-regex "CMake.FileDownload|CTestTestUpload|RunCMake.ctest_submit" Test project /usr/obj/ports/re2-20201101/build-amd64 Start 1: charclass_test 1/20 Test #1: charclass_test ... Passed0.01 sec Start 2: compile_test 2/20 Test #2: compile_test . Passed0.01 sec Start 3: filtered_re2_test 3/20 Test #3: filtered_re2_test Passed0.01 sec Start 4: mimics_pcre_test 4/20 Test #4: mimics_pcre_test . Passed0.01 sec Start 5: parse_test 5/20 Test #5: parse_test ... Passed0.01 sec Start 6: possible_match_test 6/20 Test #6: possible_match_test .. Passed6.54 sec Start 7: re2_test 7/20 Test #7: re2_test . Passed0.20 sec Start 8: re2_arg_test 8/20 Test #8: re2_arg_test . Passed0.01 sec Start 9: regexp_test 9/20 Test #9: regexp_test .. Passed0.04 sec Start 10: required_prefix_test 10/20 Test #10: required_prefix_test . Passed0.01 sec Start 11: search_test 11/20 Test #11: search_test .. Passed0.54 sec Start 12: set_test 12/20 Test #12: set_test . Passed0.01 sec Start 13: simplify_test 13/20 Test #13: simplify_test Passed0.01 sec Start 14: string_generator_test 14/20 Test #14: string_generator_test Passed0.01 sec Start 15: dfa_test 15/20 Test #15: dfa_test . Passed 48.52 sec Start 16: exhaustive1_test 16/20 Test #16: exhaustive1_test . Passed 1411.48 sec Start 17: exhaustive2_test 17/20 Test #17: exhaustive2_test . Passed 513.83 sec Start 18: exhaustive3_test 18/20 Test #18: exhaustive3_test . Passed 209.72 sec Start 19: exhaustive_test 19/20 Test #19: exhaustive_test ..***Timeout 1500.05 sec Start 20: random_test 20/20 Test #20: random_test .. Passed 178.77 sec 95% tests passed, 1 tests failed out of 20 Total Test time (real) = 3869.77 sec The following tests FAILED: 19 - exhaustive_test (Timeout) Errors while running CTest FAILED: CMakeFiles/test.util cd /usr/obj/ports/re2-20201101/build-amd64 && /usr/local/bin/ctest --force-new-ctest-process --exclude-regex "CMake.FileDownload|CTestTestUpload|RunCMake.ctest_submit" ninja: build stopped: subcommand failed. *** Error 1 in . (/usr/ports/devel/cmake/cmake.port.mk:46 'do-test': @cd /usr/obj/ports/re2-20201101/build-amd64 && exec /usr/bin/env -i LIB...) *** Error 2 in . (/usr/ports/infrastructure/mk/bsd.port.mk:2967 '/usr/obj/ports/re2-20201101/build-amd64/.test_done': @cd /usr/ports/mystuff...) *** Error 2 in /usr/ports/mystuff/textproc/re2 (/usr/ports/infrastructure/mk/bsd.port.mk:2597 'test': @lock=re2-20201101; export _LOCKS_HEL...)
Re: [New] devel/re2
Stuart Henderson writes:
> textproc would be just fine and is less full than devel
Fixed.
> : GH_ACCOUNT= google
> : GH_PROJECT= re2
> : GH_TAGNAME= 2020-11-01
> :
> : # This is needed because otherwise the package build pukes due to it
> : # not understanding the versioning convention.
> : PKGNAME=re2-${GH_TAGNAME:S,-,,g}
>
> the comment isn't needed, this situation is explicitly called out in
> packages-specs(7)
Fixed.
> :
> : SHARED_LIBS=re2 0.0
> :
> : MAINTAINER= Ashton Fagg
> :
> : # BSD
> : PERMIT_PACKAGE= Yes
> :
> : COMPILER= base-clang ports-gcc
> : MODULES=devel/cmake
> :
> : WANTLIB=${COMPILER_LIBCXX} m pthread
>
> please put wantlib in the standard place as in Makefile.template,
> when someone gets round to doing full wantlib syncs through the ports
> tree again it makes it easier to have that in the usual place
Fixed.
Update tarball attached. Sorry for the slow response.
Ash
re2.tgz
Description: Binary data
Re: [New] devel/re2
On 2020/11/13 14:57, Ashton Fagg wrote:
> Attached is a new port for re2 [1]. To quote the authors:
>
> "RE2 is a fast, safe, thread-friendly alternative to backtracking
> regular expression engines like those used in PCRE, Perl, and Python. It
> is a C++ library."
>
> NOTE: there are similarly named ports already, this is different software.
>
> Package builds and installs fine on amd64. The test suite is quite
> lengthy. On my fairly modest test machine it takes about 30-45 mins or
> so. With the default timeout settings, one of the tests does fail due to
> time out (at least on my machine), but if you increase the time limit
> it's fine. I've left them default here because I'm not sure what to do
> about it.
>
> Feedback welcomed.
>
> Ash
>
> [1]: https://github.com/google/re2
>
: CATEGORIES= devel
textproc would be just fine and is less full than devel
: GH_ACCOUNT= google
: GH_PROJECT= re2
: GH_TAGNAME= 2020-11-01
:
: # This is needed because otherwise the package build pukes due to it
: # not understanding the versioning convention.
: PKGNAME=re2-${GH_TAGNAME:S,-,,g}
the comment isn't needed, this situation is explicitly called out in
packages-specs(7)
:
: SHARED_LIBS=re2 0.0
:
: MAINTAINER= Ashton Fagg
:
: # BSD
: PERMIT_PACKAGE= Yes
:
: COMPILER= base-clang ports-gcc
: MODULES=devel/cmake
:
: WANTLIB=${COMPILER_LIBCXX} m pthread
please put wantlib in the standard place as in Makefile.template,
when someone gets round to doing full wantlib syncs through the ports
tree again it makes it easier to have that in the usual place
:
: CONFIGURE_ARGS+=-DBUILD_SHARED_LIBS=on
:
: .include
:
otherwise looks good.
Re: [New] devel/re2
пт, 13 нояб. 2020 г. в 22:58, Ashton Fagg : > > Attached is a new port for re2 [1]. To quote the authors: > > "RE2 is a fast, safe, thread-friendly alternative to backtracking > regular expression engines like those used in PCRE, Perl, and Python. It > is a C++ library." > > NOTE: there are similarly named ports already, this is different software. > > Package builds and installs fine on amd64. The test suite is quite > lengthy. On my fairly modest test machine it takes about 30-45 mins or > so. With the default timeout settings, one of the tests does fail due to > time out (at least on my machine), but if you increase the time limit > it's fine. I've left them default here because I'm not sure what to do > about it. > > Feedback welcomed. This is almost identical to the port I've recently added in openbsd-wip as a part of some ongoing work, so you have my okay here. :) -- WBR, Vadim Zhukov
[New] devel/re2
Attached is a new port for re2 [1]. To quote the authors: "RE2 is a fast, safe, thread-friendly alternative to backtracking regular expression engines like those used in PCRE, Perl, and Python. It is a C++ library." NOTE: there are similarly named ports already, this is different software. Package builds and installs fine on amd64. The test suite is quite lengthy. On my fairly modest test machine it takes about 30-45 mins or so. With the default timeout settings, one of the tests does fail due to time out (at least on my machine), but if you increase the time limit it's fine. I've left them default here because I'm not sure what to do about it. Feedback welcomed. Ash [1]: https://github.com/google/re2 re2.tgz Description: Binary data
