msg88689 - (view) |
Author: Sridhar Ratnakumar (srid) |
Date: 2009-06-02 00:00 |
$ make frameworkinstall DESTDIR=[...]/py3_1rc1-macosx-apy31-rrun/image
running install_scripts
[...]
copying build/scripts-3.1/2to3 ->
//Users/apy/rrun/build/activepython-svn-trunk/build/py3_1rc1-macosx-apy31-rrun/image/Library/Frameworks/Python.framework/Versions/3.1/bin
error:
//Users/apy/rrun/build/activepython-svn-trunk/build/py3_1rc1-macosx-apy31-rrun/image/Library/Frameworks/Python.framework/Versions/3.1/bin/2to3:
Too many levels of symbolic links
make: *** [sharedinstall] Error 1
!
!
this appears to be a regression in 3.1rc1 (used to work with 3.1b2).
need to investigate further. will update this bug if I find anything
useful. (quick observation .. 'install_scripts' seem to be run twice)
|
msg88690 - (view) |
Author: Sridhar Ratnakumar (srid) |
Date: 2009-06-02 00:01 |
To explain further, the following section of script is run *twice* ..
and thus explains the error "Too many levels of symbolic links"
mv
"[...]/image/Library/Frameworks/Python.framework/Versions/3.1/bin/2to3"
"[...]]/image/Library/Frameworks/Python.framework/Versions/3.1/bin/2to3-3.1"
ln -sf
"[...]/image/Library/Frameworks/Python.framework/Versions/3.1/bin/2to3-3.1"
"[...]/image/Library/Frameworks/Python.framework/Versions/3.1/bin/2to3"
The entire log file is here: http://double.activestate.com/t/apy31-anole.log
On 09-06-01 04:46 PM, Sridhar Ratnakumar wrote:
> >
> > make: [image_python] PythonLauncher build attempt 1 of 2 (see bug 44709)
> > make: [image_python] running 'cd build/py3_1rc1-macosx-apy31-rrun/python
> > && make frameworkinstallapps
> >
DESTDIR=/Users/apy/rrun/build/activepython-svn-trunk/build/py3_1rc1-macosx-apy31-rrun/image'
> > [...]
> > make: [image_python] running 'cd build/py3_1rc1-macosx-apy31-rrun/python
> > && make frameworkinstall frameworkinstallextras
> >
DESTDIR=/Users/apy/rrun/build/activepython-svn-trunk/build/py3_1rc1-macosx-apy31-rrun/image'
> > [...]
> >
> >
> > Note: the second make attempt uses 'frameworkinstallextras' as the extra
> > argument.
> >
> > .
> > .
> >
> > This appears to be breaking 3.1rc1 on Mac ("Too many levels of symbolic
> > links"):
> >
> > running install_scripts
> > [...]
> > copying build/scripts-3.1/2to3 ->
> >
//Users/apy/rrun/build/activepython-svn-trunk/build/py3_1rc1-macosx-apy31-rrun/image/Library/Frameworks/Python.framework/Versions/3.1/bin
> > error:
> >
//Users/apy/rrun/build/activepython-svn-trunk/build/py3_1rc1-macosx-apy31-rrun/image/Library/Frameworks/Python.framework/Versions/3.1/bin/2to3:
> > Too many levels of symbolic links
> > make: *** [sharedinstall] Error 1
|
msg88691 - (view) |
Author: Sridhar Ratnakumar (srid) |
Date: 2009-06-02 00:02 |
Tarek, are you perchance aware of any change that went into distutils
since the last beta release that could have influenced the make behavior?
|
msg88693 - (view) |
Author: Tarek Ziadé (tarek) * |
Date: 2009-06-02 00:10 |
No, but I am running frameworkinstall right now under Mac to see if I
can reproduce the problem
|
msg88694 - (view) |
Author: Sridhar Ratnakumar (srid) |
Date: 2009-06-02 00:12 |
To help with the repro, may I suggest running the following in the same
order:
make frameworkinstallframework DESTDIR=image
make frameworkinstallapps DESTDIR=image
make frameworkinstall frameworkinstallextras DESTDIR=image
|
msg88696 - (view) |
Author: Tarek Ziadé (tarek) * |
Date: 2009-06-02 00:24 |
I'll try tomorrow asap (sorry it's 2:30 am now here)
|
msg88697 - (view) |
Author: Sridhar Ratnakumar (srid) |
Date: 2009-06-02 00:27 |
also related to issue 5756
|
msg88721 - (view) |
Author: Ned Deily (ned.deily) * |
Date: 2009-06-02 07:18 |
The URL you provide for the log doesn't seem to be accessible externally
so it is difficult to guess exactly what was being done. However, there
were changes between 3.1b2 and 3.1rc1 to the top-level Makefile
(Makefile.pre.in) and to the Mac/Makefile.in affecting framework builds
and installs.
|
msg88726 - (view) |
Author: Tarek Ziadé (tarek) * |
Date: 2009-06-02 09:01 |
Sorry I can't reproduce it,
can you tell me how you run configure precisely ?
eg the whole set of commands after a fresh checkout
|
msg88728 - (view) |
Author: Ronald Oussoren (ronaldoussoren) * |
Date: 2009-06-02 09:07 |
I also cannot reproduce this, using both the py3k branch and the r31rc1
tag.
My build procedure:
* cd r31rc1
* mkdir build
* cd build
* ../configure --enable-framework --enable-universalsdk
* make
* make install DESTDIR=$PWD/image
The resulting framework in the $DESTDIR seems to be complete, and I
don't get errors while building it.
|
msg88731 - (view) |
Author: Ronald Oussoren (ronaldoussoren) * |
Date: 2009-06-02 09:17 |
Now I'm really confused, I tried to install a second time and this time I
do get an error.
Time to start hunting down an issue...
|
msg88732 - (view) |
Author: Tarek Ziadé (tarek) * |
Date: 2009-06-02 09:19 |
Ronald, did you try Sridar precise sequence after Pyhon was built ?
make frameworkinstallframework DESTDIR=image
make frameworkinstallapps DESTDIR=image
make frameworkinstall frameworkinstallextras DESTDIR=image
|
msg88733 - (view) |
Author: Ronald Oussoren (ronaldoussoren) * |
Date: 2009-06-02 09:21 |
Manually running "make frameworkinstallframework" (and the other ones) is
not supported, those are internal makefile targets that are are used to
implement "make install".
"make framworkinstall" is supported as an alias for "make install" because
the former was the only way to install a framework for a long time.
|
msg88736 - (view) |
Author: Ronald Oussoren (ronaldoussoren) * |
Date: 2009-06-02 10:12 |
I found the issue.
In r72899 antoine.pitrou checked in a change that adds these two lines
to the bininstall target in Makefile.pre.in:
-rm -f $(DESTDIR)$(LIBPC)/python3.pc
(cd $(DESTDIR)$(LIBPC); $(LN) -s python-$(VERSION).pc python3.pc)
This exposes a bug in the configure script, which causes target
'bininstall' to run before 'libainstall' when your doing a framework
install. The latter target is the one that creates python-
$(VERSION).pc, which is why we get build failures.
What I don't understand is why I've been able to install frameworks all
along and didn't notice this before.
I'm working on a patch, and will check that in after I've full tested
it.
The important bit of the patch is this change to configure.in:
- FRAMEWORKALTINSTALLFIRST="frameworkinstallstructure
bininstall maninstall"
+ FRAMEWORKALTINSTALLFIRST="frameworkinstallstructure "
|
msg88737 - (view) |
Author: Ronald Oussoren (ronaldoussoren) * |
Date: 2009-06-02 10:56 |
This should be fixed in r73142, please test.
|
msg88784 - (view) |
Author: Ned Deily (ned.deily) * |
Date: 2009-06-03 01:57 |
That seems to work. However, while testing that, I found another, minor
problem in that same area. In Makefile.pre.in bininstall target just
before the lines you cite, there is also this added in r71935:
-rm -f $(DESTDIR)$(BINDIR)/python3-config
(cd $(DESTDIR)$(BINDIR); $(LN) -s python$(VERSION)-config python3-
config)
Prior to that, though, the Mac Makefile target install_versionedtools is
mucking in the same area:
if [ ! -h "$(DESTDIR)$(prefix)/bin/python3-config" ]; then \
mv "$(DESTDIR)$(prefix)/bin/python3-config"
"$(DESTDIR)$(prefix)/bin/python$(VERSION)-config" ;\
ln -sf "python$(VERSION)-config" "$(DESTDIR)$(prefix)/bin/python3-
config" ; \
fi
and, in fact, the mv fails with an error because (1) the python3-config
hasn't been created yet and (2) the test is incorrect: ! -h is also true
if the file doesn't exist. With the code in bininstall, I think these
lines can be safely removed from install_versionedtools.
|
msg88852 - (view) |
Author: Sridhar Ratnakumar (srid) |
Date: 2009-06-03 23:38 |
Ronald, a quick note - your fix unfortunately did not solve the issue.
[quote]'To help with the repro, may I suggest running the following in
the same
order:
make frameworkinstallframework DESTDIR=image
make frameworkinstallapps DESTDIR=image
make frameworkinstall frameworkinstallextras DESTDIR=image
'[endquote[
|
msg88853 - (view) |
Author: Ronald Oussoren (ronaldoussoren) * |
Date: 2009-06-03 23:41 |
Don't do that, use 'make install DESTDIR=$PWD/image'. As I mentioned
earlier 'make frameworkinstallframework' is an internal step that
shouldn't be used on its own.
|
msg89747 - (view) |
Author: Sridhar Ratnakumar (srid) |
Date: 2009-06-26 23:42 |
Ok, I am no longer relying on internal targets .. and this problem is
fixed for me. It needs to be closed, correct?
|
|
Date |
User |
Action |
Args |
2022-04-11 14:56:49 | admin | set | github: 50420 |
2009-06-26 23:42:36 | srid | set | status: open -> closed
messages:
+ msg89747 |
2009-06-04 06:32:14 | tarek | set | assignee: tarek -> |
2009-06-03 23:41:23 | ronaldoussoren | set | messages:
+ msg88853 |
2009-06-03 23:38:05 | srid | set | messages:
+ msg88852 |
2009-06-03 01:57:58 | ned.deily | set | status: pending -> open
messages:
+ msg88784 |
2009-06-02 10:56:39 | ronaldoussoren | set | status: open -> pending resolution: fixed messages:
+ msg88737
|
2009-06-02 10:12:49 | ronaldoussoren | set | nosy:
ronaldoussoren, tarek, ned.deily, srid messages:
+ msg88736 components:
+ macOS, - Distutils, 2to3 (2.x to 3.x conversion tool) |
2009-06-02 09:21:40 | ronaldoussoren | set | messages:
+ msg88733 |
2009-06-02 09:19:44 | tarek | set | messages:
+ msg88732 |
2009-06-02 09:17:49 | ronaldoussoren | set | messages:
+ msg88731 |
2009-06-02 09:07:35 | ronaldoussoren | set | messages:
+ msg88728 |
2009-06-02 09:01:01 | tarek | set | messages:
+ msg88726 |
2009-06-02 07:19:18 | ned.deily | set | nosy:
+ ronaldoussoren, ned.deily messages:
+ msg88721
|
2009-06-02 00:27:33 | srid | set | messages:
+ msg88697 components:
+ Build, 2to3 (2.x to 3.x conversion tool) |
2009-06-02 00:24:30 | tarek | set | messages:
+ msg88696 |
2009-06-02 00:12:11 | srid | set | messages:
+ msg88694 |
2009-06-02 00:10:13 | tarek | set | messages:
+ msg88693 |
2009-06-02 00:02:32 | srid | set | messages:
+ msg88691 |
2009-06-02 00:01:28 | srid | set | messages:
+ msg88690 |
2009-06-02 00:00:48 | srid | set | files:
- apy31-anole.log |
2009-06-02 00:00:39 | srid | create | |