You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by Noorul Islam K M <no...@collab.net> on 2011/06/24 06:41:50 UTC
Implement support for 'make check' variants
Daniel came up with idea of extending Makefile.in to support 'make
check' variants.
1. 'make check' with svnserve -T
2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
(1.4, 1.5, 1.6, 1.7))
3. 'make check' with SVNPathAuthz short_circuit. As of now this is
supported by stsp's Makefile.svn. So this will be bringing that into
Makefile.in
So, this would include the following tasks.
* Extending Makefile.in to take optional arguments.
* Propagating that through to the tests themselves.
* Extending test suite to take additional options accordingly.
* Filing issues and/or fixing any bugs that surface.
I am planning to implement his idea. I hope this is a welcome
change. Posting it here to get feedback before starting to work on this.
Thanks and Regards
Noorul
Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Noorul Islam K M wrote on Tue, Jul 05, 2011 at 09:55:08 +0530:
> Daniel Shahaf <d....@daniel.shahaf.name> writes:
>
> > Noorul Islam K M wrote on Mon, Jul 04, 2011 at 17:52:53 +0530:
> >
> >> Noorul Islam K M <no...@collab.net> writes:
> >>
> >> An update
> >>
> >> > Daniel came up with idea of extending Makefile.in to support 'make
> >> > check' variants.
> >> >
> >> > 1. 'make check' with svnserve -T
> >>
> >> This was already implemented.
> >> Completed testing using "make svnserveautocheck THREADED=1". I could not
> >> see any errors.
> >>
> >> > 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
> >> > (1.4, 1.5, 1.6, 1.7))
> >>
> >> This was already implemented.
> >> make check SERVER_MINOR_VERSION=3.
> >> I could see some failures. Are all the tests supposed to pass with this
> >> kind of testing?
> >>
> >
> > If you see failures please report them on separate threads; they might
> > be shortcomings in the test suite, but let's confirm that.
> >
>
> I will do this.
>
> > By the way: according to my reading of the test suite's code, you should
> > have gotten an error message if you'd passed SERVER_MINOR_VERSION=3
> > (rather than 4 or bigger)! How come you didn't trigger the check in
> > svntest/main.py _parse_options()?
> >
>
> _parse_options() will not come into picture if we use 'make check'
> because we are directly calling svntest.main.execute_tests() from
> run_tests.py.
>
> I think we should patch run_tests.py to display error if the user passes
> values that is not in range 4 to 6.
>
> One more thing that I observed is that we are not supporting
> --pre-1.4-compatible option even though this one is available as an
> option to svnadmin command. See the snippet below. Is this intentional?
>
Feel free to add it and see what breaks...
(again, the SERVER_MINOR_VERSION options are used for two things, for
passing the --pre-1.x-compatible flag and for adjusting tests'
expectations when a 1.x.0 server is started *manually* to run the tests
against)
> <code>
> if options.server_minor_version < 5:
> opts += ("--pre-1.5-compatible",)
> elif options.server_minor_version < 6:
> opts += ("--pre-1.6-compatible",)
> elif options.server_minor_version < 7:
> opts += ("--pre-1.7-compatible",)
> if options.fs_type is not None:
> opts += ("--fs-type=" + options.fs_type,)
> exit_code, stdout, stderr = run_command(svnadmin_binary, 1, 0, "create",
> path, *opts)
> </code>
>
> Thanks and Regards
> Noorul
Re: Implement support for 'make check' variants
Posted by Noorul Islam K M <no...@collab.net>.
Daniel Shahaf <d....@daniel.shahaf.name> writes:
> Noorul Islam K M wrote on Mon, Jul 04, 2011 at 17:52:53 +0530:
>
>> Noorul Islam K M <no...@collab.net> writes:
>>
>> An update
>>
>> > Daniel came up with idea of extending Makefile.in to support 'make
>> > check' variants.
>> >
>> > 1. 'make check' with svnserve -T
>>
>> This was already implemented.
>> Completed testing using "make svnserveautocheck THREADED=1". I could not
>> see any errors.
>>
>> > 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
>> > (1.4, 1.5, 1.6, 1.7))
>>
>> This was already implemented.
>> make check SERVER_MINOR_VERSION=3.
>> I could see some failures. Are all the tests supposed to pass with this
>> kind of testing?
>>
>
> If you see failures please report them on separate threads; they might
> be shortcomings in the test suite, but let's confirm that.
>
I will do this.
> By the way: according to my reading of the test suite's code, you should
> have gotten an error message if you'd passed SERVER_MINOR_VERSION=3
> (rather than 4 or bigger)! How come you didn't trigger the check in
> svntest/main.py _parse_options()?
>
_parse_options() will not come into picture if we use 'make check'
because we are directly calling svntest.main.execute_tests() from
run_tests.py.
I think we should patch run_tests.py to display error if the user passes
values that is not in range 4 to 6.
One more thing that I observed is that we are not supporting
--pre-1.4-compatible option even though this one is available as an
option to svnadmin command. See the snippet below. Is this intentional?
<code>
if options.server_minor_version < 5:
opts += ("--pre-1.5-compatible",)
elif options.server_minor_version < 6:
opts += ("--pre-1.6-compatible",)
elif options.server_minor_version < 7:
opts += ("--pre-1.7-compatible",)
if options.fs_type is not None:
opts += ("--fs-type=" + options.fs_type,)
exit_code, stdout, stderr = run_command(svnadmin_binary, 1, 0, "create",
path, *opts)
</code>
Thanks and Regards
Noorul
Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Noorul Islam K M wrote on Mon, Jul 04, 2011 at 17:52:53 +0530:
> Noorul Islam K M <no...@collab.net> writes:
>
> An update
>
> > Daniel came up with idea of extending Makefile.in to support 'make
> > check' variants.
> >
> > 1. 'make check' with svnserve -T
>
> This was already implemented.
> Completed testing using "make svnserveautocheck THREADED=1". I could not
> see any errors.
>
> > 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
> > (1.4, 1.5, 1.6, 1.7))
>
> This was already implemented.
> make check SERVER_MINOR_VERSION=3.
> I could see some failures. Are all the tests supposed to pass with this
> kind of testing?
>
If you see failures please report them on separate threads; they might
be shortcomings in the test suite, but let's confirm that.
By the way: according to my reading of the test suite's code, you should
have gotten an error message if you'd passed SERVER_MINOR_VERSION=3
(rather than 4 or bigger)! How come you didn't trigger the check in
svntest/main.py _parse_options()?
> > 3. 'make check' with SVNPathAuthz short_circuit. As of now this is
> > supported by stsp's Makefile.svn. So this will be bringing that into
> > Makefile.in
> >
>
> Implemented and tested. No issues found while testing.
>
Thanks.
> Thanks and Regards
> Noorul
Re: Implement support for 'make check' variants
Posted by Noorul Islam K M <no...@collab.net>.
Noorul Islam K M <no...@collab.net> writes:
An update
> Daniel came up with idea of extending Makefile.in to support 'make
> check' variants.
>
> 1. 'make check' with svnserve -T
This was already implemented.
Completed testing using "make svnserveautocheck THREADED=1". I could not
see any errors.
> 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
> (1.4, 1.5, 1.6, 1.7))
This was already implemented.
make check SERVER_MINOR_VERSION=3.
I could see some failures. Are all the tests supposed to pass with this
kind of testing?
> 3. 'make check' with SVNPathAuthz short_circuit. As of now this is
> supported by stsp's Makefile.svn. So this will be bringing that into
> Makefile.in
>
Implemented and tested. No issues found while testing.
Thanks and Regards
Noorul
Re: [PATCH] Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
r1141292. (See also r1141291.) Thanks!
Noorul Islam K M wrote on Wed, Jun 29, 2011 at 15:58:04 +0530:
> Noorul Islam K M <no...@collab.net> writes:
>
> > Daniel Shahaf <d....@daniel.shahaf.name> writes:
> >
> >> Reads fine.
> >>
> >> I'd like to test this patch (perhaps later this week?).
> >>
> >
> > I tested with short_circuit configuration. I did not see any errors.
> >
> >> Would
> >> make davautocheck ENVVAR=short_circuit
> >> be a better interface?
> >>
> >
> > I made it this way.
> >
> > make davautocheck SVN_PATH_AUTHZ=short_circuit
> >
> > Attached is the patch.
> >
> > Log
> >
> > [[[
> >
> > Allow davautocheck to run optionally by setting path based authz as
> > "short_circuit".
> >
> > * Makefile.in
> > (davautocheck): Document environment parameter.
> >
> > * subversion/tests/cmdline/davautocheck.sh:
> > Accept and use $SVN_PATH_AUTHZ environment parameter.
> > ]]]
> >
>
> Please ignore the last patch. I missed to replace a text. Attached is
> the correct one. Also pasting the log message once again.
>
> Log
>
> [[[
>
> Allow davautocheck to run optionally by setting path based authz as
> "short_circuit".
>
> * Makefile.in
> (davautocheck): Document environment parameter.
>
> * subversion/tests/cmdline/davautocheck.sh:
> Accept and use $SVN_PATH_AUTHZ environment parameter.
>
> Patch by: Noorul Islam K M <noorul{_AT_}collab.net>
> Suggested by: danielsh
>
> ]]]
>
> Index: Makefile.in
> ===================================================================
> --- Makefile.in (revision 1141012)
> +++ Makefile.in (working copy)
> @@ -501,7 +501,7 @@
> # Automatically configure and run Apache httpd on a random port, and then
> # run make check.
> davautocheck: bin $(TEST_DEPS) @BDB_TEST_DEPS@ apache-mod
> - @# Takes MODULE_PATH and USE_HTTPV1 in the environment.
> + @# Takes MODULE_PATH, USE_HTTPV1 and SVN_PATH_AUTHZ in the environment.
> @APXS=$(APXS) bash $(top_srcdir)/subversion/tests/cmdline/davautocheck.sh
>
> # First, run:
> Index: subversion/tests/cmdline/davautocheck.sh
> ===================================================================
> --- subversion/tests/cmdline/davautocheck.sh (revision 1141012)
> +++ subversion/tests/cmdline/davautocheck.sh (working copy)
> @@ -67,6 +67,9 @@
> #
> # To prevent the server from advertising httpv2, pass USE_HTTPV1 in
> # the environment.
> +#
> +# To use value for "SVNPathAuthz" directive set SVN_PATH_AUTHZ with
> +# appropriate value in the environment.
> #
> # Passing --no-tests as argv[1] will have the script start a server
> # but not run any tests.
> @@ -160,6 +163,12 @@
> ADVERTISE_V2_PROTOCOL=off
> fi
>
> +# Pick up $SVN_PATH_AUTHZ
> +SVN_PATH_AUTHZ_LINE=""
> +if [ ${SVN_PATH_AUTHZ:+set} ]; then
> + SVN_PATH_AUTHZ_LINE="SVNPathAuthz ${SVN_PATH_AUTHZ}"
> +fi
> +
> # Find the source and build directories. The build dir can be found if it is
> # the current working dir or the source dir.
> pushd ${SCRIPTDIR}/../../../ > /dev/null
> @@ -326,6 +335,7 @@
> AuthUserFile $HTTPD_USERS
> Require valid-user
> SVNAdvertiseV2Protocol ${ADVERTISE_V2_PROTOCOL}
> + ${SVN_PATH_AUTHZ_LINE}
> </Location>
> <Location /svn-test-work/local_tmp/repos>
> DAV svn
> @@ -336,6 +346,7 @@
> AuthUserFile $HTTPD_USERS
> Require valid-user
> SVNAdvertiseV2Protocol ${ADVERTISE_V2_PROTOCOL}
> + ${SVN_PATH_AUTHZ_LINE}
> </Location>
> RedirectMatch permanent ^/svn-test-work/repositories/REDIRECT-PERM-(.*)\$ /svn-test-work/repositories/\$1
> RedirectMatch ^/svn-test-work/repositories/REDIRECT-TEMP-(.*)\$ /svn-test-work/repositories/\$1
Re: [PATCH] Re: Implement support for 'make check' variants
Posted by Noorul Islam K M <no...@collab.net>.
Noorul Islam K M <no...@collab.net> writes:
> Daniel Shahaf <d....@daniel.shahaf.name> writes:
>
>> Reads fine.
>>
>> I'd like to test this patch (perhaps later this week?).
>>
>
> I tested with short_circuit configuration. I did not see any errors.
>
>> Would
>> make davautocheck ENVVAR=short_circuit
>> be a better interface?
>>
>
> I made it this way.
>
> make davautocheck SVN_PATH_AUTHZ=short_circuit
>
> Attached is the patch.
>
> Log
>
> [[[
>
> Allow davautocheck to run optionally by setting path based authz as
> "short_circuit".
>
> * Makefile.in
> (davautocheck): Document environment parameter.
>
> * subversion/tests/cmdline/davautocheck.sh:
> Accept and use $SVN_PATH_AUTHZ environment parameter.
> ]]]
>
Please ignore the last patch. I missed to replace a text. Attached is
the correct one. Also pasting the log message once again.
Log
[[[
Allow davautocheck to run optionally by setting path based authz as
"short_circuit".
* Makefile.in
(davautocheck): Document environment parameter.
* subversion/tests/cmdline/davautocheck.sh:
Accept and use $SVN_PATH_AUTHZ environment parameter.
Patch by: Noorul Islam K M <noorul{_AT_}collab.net>
Suggested by: danielsh
]]]
Re: [PATCH] Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Noorul Islam K M wrote on Wed, Jun 29, 2011 at 15:43:04 +0530:
> Daniel Shahaf <d....@daniel.shahaf.name> writes:
>
> > Reads fine.
> >
> > I'd like to test this patch (perhaps later this week?).
> >
>
> I tested with short_circuit configuration. I did not see any errors.
>
In these kind of patches when I say "test" I mean "verify that the
configuration intended to take effect, does". Thanks for reporting your
results.
> > Would
> > make davautocheck ENVVAR=short_circuit
> > be a better interface?
> >
>
> I made it this way.
>
> make davautocheck SVN_PATH_AUTHZ=short_circuit
>
I take it you agree with me then :-)
Anyway, +1 to the interface. I'll probably commit this patch (i.e., its
followup that you've posted) once I've tested it. Thanks!
> Attached is the patch.
>
> Log
>
> [[[
>
> Allow davautocheck to run optionally by setting path based authz as
> "short_circuit".
>
> * Makefile.in
> (davautocheck): Document environment parameter.
>
> * subversion/tests/cmdline/davautocheck.sh:
> Accept and use $SVN_PATH_AUTHZ environment parameter.
> ]]]
>
> Thanks and Regards
> Noorul
>
> Index: Makefile.in
> ===================================================================
> --- Makefile.in (revision 1141012)
> +++ Makefile.in (working copy)
> @@ -501,7 +501,7 @@
> # Automatically configure and run Apache httpd on a random port, and then
> # run make check.
> davautocheck: bin $(TEST_DEPS) @BDB_TEST_DEPS@ apache-mod
> - @# Takes MODULE_PATH and USE_HTTPV1 in the environment.
> + @# Takes MODULE_PATH, USE_HTTPV1 and SVN_PATH_AUTHZ in the environment.
> @APXS=$(APXS) bash $(top_srcdir)/subversion/tests/cmdline/davautocheck.sh
>
> # First, run:
> Index: subversion/tests/cmdline/davautocheck.sh
> ===================================================================
> --- subversion/tests/cmdline/davautocheck.sh (revision 1141012)
> +++ subversion/tests/cmdline/davautocheck.sh (working copy)
> @@ -67,6 +67,9 @@
> #
> # To prevent the server from advertising httpv2, pass USE_HTTPV1 in
> # the environment.
> +#
> +# To use value for "SVNPathAuthz" directive set SVN_PATH_AUTHZ with
> +# appropriate value in the environment.
> #
> # Passing --no-tests as argv[1] will have the script start a server
> # but not run any tests.
> @@ -160,6 +163,12 @@
> ADVERTISE_V2_PROTOCOL=off
> fi
>
> +# Pick up $USE_AUTHZ_SHORT_CIRCUIT
> +SVN_PATH_AUTHZ_LINE=""
> +if [ ${SVN_PATH_AUTHZ:+set} ]; then
> + SVN_PATH_AUTHZ_LINE="SVNPathAuthz ${SVN_PATH_AUTHZ}"
> +fi
> +
> # Find the source and build directories. The build dir can be found if it is
> # the current working dir or the source dir.
> pushd ${SCRIPTDIR}/../../../ > /dev/null
> @@ -326,6 +335,7 @@
> AuthUserFile $HTTPD_USERS
> Require valid-user
> SVNAdvertiseV2Protocol ${ADVERTISE_V2_PROTOCOL}
> + ${SVN_PATH_AUTHZ_LINE}
> </Location>
> <Location /svn-test-work/local_tmp/repos>
> DAV svn
> @@ -336,6 +346,7 @@
> AuthUserFile $HTTPD_USERS
> Require valid-user
> SVNAdvertiseV2Protocol ${ADVERTISE_V2_PROTOCOL}
> + ${SVN_PATH_AUTHZ_LINE}
> </Location>
> RedirectMatch permanent ^/svn-test-work/repositories/REDIRECT-PERM-(.*)\$ /svn-test-work/repositories/\$1
> RedirectMatch ^/svn-test-work/repositories/REDIRECT-TEMP-(.*)\$ /svn-test-work/repositories/\$1
Re: [PATCH] Re: Implement support for 'make check' variants
Posted by Noorul Islam K M <no...@collab.net>.
Daniel Shahaf <d....@daniel.shahaf.name> writes:
> Reads fine.
>
> I'd like to test this patch (perhaps later this week?).
>
I tested with short_circuit configuration. I did not see any errors.
> Would
> make davautocheck ENVVAR=short_circuit
> be a better interface?
>
I made it this way.
make davautocheck SVN_PATH_AUTHZ=short_circuit
Attached is the patch.
Log
[[[
Allow davautocheck to run optionally by setting path based authz as
"short_circuit".
* Makefile.in
(davautocheck): Document environment parameter.
* subversion/tests/cmdline/davautocheck.sh:
Accept and use $SVN_PATH_AUTHZ environment parameter.
]]]
Thanks and Regards
Noorul
Re: [PATCH] Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Reads fine.
I'd like to test this patch (perhaps later this week?).
Would
make davautocheck ENVVAR=short_circuit
be a better interface?
Noorul Islam K M wrote on Tue, Jun 28, 2011 at 12:03:36 +0530:
> Index: Makefile.in
> ===================================================================
> --- Makefile.in (revision 1140431)
> +++ Makefile.in (working copy)
> @@ -501,7 +501,8 @@
> # Automatically configure and run Apache httpd on a random port, and then
> # run make check.
> davautocheck: bin $(TEST_DEPS) @BDB_TEST_DEPS@ apache-mod
> - @# Takes MODULE_PATH and USE_HTTPV1 in the environment.
> + @# Takes MODULE_PATH, USE_HTTPV1 and USE_AUTHZ_SHORT_CIRCUIT in the
> + @# environment.
> @APXS=$(APXS) bash $(top_srcdir)/subversion/tests/cmdline/davautocheck.sh
>
> # First, run:
> Index: subversion/tests/cmdline/davautocheck.sh
> ===================================================================
> --- subversion/tests/cmdline/davautocheck.sh (revision 1140431)
> +++ subversion/tests/cmdline/davautocheck.sh (working copy)
> @@ -67,6 +67,9 @@
> #
> # To prevent the server from advertising httpv2, pass USE_HTTPV1 in
> # the environment.
> +#
> +# To use "short_circuit" value for "SvnPathAuthz" directive set
> +# USE_AUTHZ_SHORT_CIRCUIT=yes in the environment.
> #
> # Passing --no-tests as argv[1] will have the script start a server
> # but not run any tests.
> @@ -160,6 +163,12 @@
> ADVERTISE_V2_PROTOCOL=off
> fi
>
> +# Pick up $USE_AUTHZ_SHORT_CIRCUIT
> +SVN_PATH_AUTHZ=""
> +if [ ${USE_AUTHZ_SHORT_CIRCUIT:+set} ]; then
> + SVN_PATH_AUTHZ="SVNPathAuthz short_circuit"
> +fi
[PATCH] Re: Implement support for 'make check' variants
Posted by Noorul Islam K M <no...@collab.net>.
Noorul Islam K M <no...@collab.net> writes:
> Daniel came up with idea of extending Makefile.in to support 'make
> check' variants.
>
> 1. 'make check' with svnserve -T
> 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
> (1.4, 1.5, 1.6, 1.7))
> 3. 'make check' with SVNPathAuthz short_circuit. As of now this is
> supported by stsp's Makefile.svn. So this will be bringing that into
> Makefile.in
>
Attached is the patch for #3.
Log
[[[
Allow davautocheck to run optionally by setting path based authz as
"short_circuit".
* Makefile.in
(davautocheck): Document environment parameter.
* subversion/tests/cmdline/davautocheck.sh:
Accept and use $USE_AUTHZ_SHORT_CIRCUIT environment parameter.
]]]
Thanks and Regards
Noorul
Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <da...@elego.de>.
Noorul Islam K M wrote on Fri, Jun 24, 2011 at 10:11:50 +0530:
>
> Daniel came up with idea of extending Makefile.in to support 'make
> check' variants.
>
> 1. 'make check' with svnserve -T
> 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
> (1.4, 1.5, 1.6, 1.7))
> 3. 'make check' with SVNPathAuthz short_circuit. As of now this is
> supported by stsp's Makefile.svn. So this will be bringing that into
> Makefile.in
>
And in case someone wonders why: I have in the past found a bug in
the cache code via #1; we recently had CVE's that would have been
exposed by someone running #3; and #2 is a cheap (though not as good as
'svnadmin1.x create') to test running the 1.7 code on 1.6 repositories
and filesystems.
I am not suggesting to enable any of these by default.
> So, this would include the following tasks.
>
> * Extending Makefile.in to take optional arguments.
> * Propagating that through to the tests themselves.
> * Extending test suite to take additional options accordingly.
> * Filing issues and/or fixing any bugs that surface.
>
> I am planning to implement his idea. I hope this is a welcome
> change. Posting it here to get feedback before starting to work on this.
>
> Thanks and Regards
> Noorul
Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Philip Martin wrote on Fri, Jun 24, 2011 at 13:03:27 +0100:
> Daniel Shahaf <d....@daniel.shahaf.name> writes:
>
> > SERVER_MINOR_VERSION is for testing a 1.7 client with a 1.6 server.
> >
> > I'm after testing a 1.7 client with a 1.7 server using a 1.6 filesystme.
>
> As discussed on IRC, SERVER_MINOR_VERSION should already cause a 1.6
> filesystem to be used. However there is a bug somewhere in the option
> handling that causes options.server_minor_version to be of type string
> so the comparisons against integer values don't work.
>
> The problem occurs if I run:
>
> make check SERVER_MINOR_VERSION=6
>
> but if I run
>
> ./basic_checks.py --server-minor-version 6
>
> the problem is not present. I can fix the problem using the patch below,
> but I don't understand why it is necessary, and I don't know if other
> parameters are affected.
>
In r1139268 I patched build/run_tests.py to store the integer-typed
parameters --- FSFS_SHARDING and SERVER_MINOR_VERSION --- as integers
in svntest.main.options. This fixes create_repos() issue for me.
Thanks for spotting the str/int discrepancy!
> Index: subversion/tests/cmdline/svntest/main.py
> ===================================================================
> --- subversion/tests/cmdline/svntest/main.py (revision 1139223)
> +++ subversion/tests/cmdline/svntest/main.py (working copy)
> @@ -1614,6 +1614,8 @@
> else:
> parser = _create_parser()
>
> + options.server_minor_version = int(options.server_minor_version)
> +
> # parse the positional arguments (test nums, names)
> for arg in test_selection:
> appended = False
>
> --
> Philip
Re: Implement support for 'make check' variants
Posted by Philip Martin <ph...@wandisco.com>.
Daniel Shahaf <d....@daniel.shahaf.name> writes:
> SERVER_MINOR_VERSION is for testing a 1.7 client with a 1.6 server.
>
> I'm after testing a 1.7 client with a 1.7 server using a 1.6 filesystme.
As discussed on IRC, SERVER_MINOR_VERSION should already cause a 1.6
filesystem to be used. However there is a bug somewhere in the option
handling that causes options.server_minor_version to be of type string
so the comparisons against integer values don't work.
The problem occurs if I run:
make check SERVER_MINOR_VERSION=6
but if I run
./basic_checks.py --server-minor-version 6
the problem is not present. I can fix the problem using the patch below,
but I don't understand why it is necessary, and I don't know if other
parameters are affected.
Index: subversion/tests/cmdline/svntest/main.py
===================================================================
--- subversion/tests/cmdline/svntest/main.py (revision 1139223)
+++ subversion/tests/cmdline/svntest/main.py (working copy)
@@ -1614,6 +1614,8 @@
else:
parser = _create_parser()
+ options.server_minor_version = int(options.server_minor_version)
+
# parse the positional arguments (test nums, names)
for arg in test_selection:
appended = False
--
Philip
Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
Philip Martin wrote on Fri, Jun 24, 2011 at 11:47:44 +0100:
> Noorul Islam K M <no...@collab.net> writes:
>
> > Daniel came up with idea of extending Makefile.in to support 'make
> > check' variants.
> >
> > 1. 'make check' with svnserve -T
>
> > 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
> > (1.4, 1.5, 1.6, 1.7))
> > 3. 'make check' with SVNPathAuthz short_circuit. As of now this is
> > supported by stsp's Makefile.svn. So this will be bringing that into
> > Makefile.in
> >
> > So, this would include the following tasks.
> >
> > * Extending Makefile.in to take optional arguments.
> > * Propagating that through to the tests themselves.
> > * Extending test suite to take additional options accordingly.
> > * Filing issues and/or fixing any bugs that surface.
> >
> > I am planning to implement his idea. I hope this is a welcome
> > change. Posting it here to get feedback before starting to work on this.
>
> I don't understand what you intend. I agree that 3 is new, but for 1
> and 2 I can do:
>
> make svnserveautocheck THREADED=1
> make check SERVER_MINOR_VERSION=N
>
> Are you plan to change/extend those options?
SERVER_MINOR_VERSION is for testing a 1.7 client with a 1.6 server.
I'm after testing a 1.7 client with a 1.7 server using a 1.6 filesystme.
(Also, SERVER_MINOR_VERSION needs you to start the 1.6 server manually)
Re: Implement support for 'make check' variants
Posted by Daniel Shahaf <d....@daniel.shahaf.name>.
For THREADED, no I don't plan to change it, I simply forgot that I had
already added it.
Philip Martin wrote on Fri, Jun 24, 2011 at 11:47:44 +0100:
> Noorul Islam K M <no...@collab.net> writes:
>
> > Daniel came up with idea of extending Makefile.in to support 'make
> > check' variants.
> >
> > 1. 'make check' with svnserve -T
>
> > 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
> > (1.4, 1.5, 1.6, 1.7))
> > 3. 'make check' with SVNPathAuthz short_circuit. As of now this is
> > supported by stsp's Makefile.svn. So this will be bringing that into
> > Makefile.in
> >
> > So, this would include the following tasks.
> >
> > * Extending Makefile.in to take optional arguments.
> > * Propagating that through to the tests themselves.
> > * Extending test suite to take additional options accordingly.
> > * Filing issues and/or fixing any bugs that surface.
> >
> > I am planning to implement his idea. I hope this is a welcome
> > change. Posting it here to get feedback before starting to work on this.
>
> I don't understand what you intend. I agree that 3 is new, but for 1
> and 2 I can do:
>
> make svnserveautocheck THREADED=1
> make check SERVER_MINOR_VERSION=N
>
> Are you plan to change/extend those options?
>
> --
> Philip
Re: Implement support for 'make check' variants
Posted by Philip Martin <ph...@wandisco.com>.
Noorul Islam K M <no...@collab.net> writes:
> Daniel came up with idea of extending Makefile.in to support 'make
> check' variants.
>
> 1. 'make check' with svnserve -T
> 2. 'make check' with svnadmin create --pre-1.x-compatible (where x in
> (1.4, 1.5, 1.6, 1.7))
> 3. 'make check' with SVNPathAuthz short_circuit. As of now this is
> supported by stsp's Makefile.svn. So this will be bringing that into
> Makefile.in
>
> So, this would include the following tasks.
>
> * Extending Makefile.in to take optional arguments.
> * Propagating that through to the tests themselves.
> * Extending test suite to take additional options accordingly.
> * Filing issues and/or fixing any bugs that surface.
>
> I am planning to implement his idea. I hope this is a welcome
> change. Posting it here to get feedback before starting to work on this.
I don't understand what you intend. I agree that 3 is new, but for 1
and 2 I can do:
make svnserveautocheck THREADED=1
make check SERVER_MINOR_VERSION=N
Are you plan to change/extend those options?
--
Philip