| 1 |
On Sat, Sep 14, 2013 at 6:38 PM, Michał Górny <mgorny@g.o> wrote: |
| 2 |
> Dnia 2013-09-14, o godz. 18:16:33 |
| 3 |
> Mike Gilbert <floppym@g.o> napisał(a): |
| 4 |
> |
| 5 |
>> On Fri, Sep 13, 2013 at 2:58 PM, Michał Górny <mgorny@g.o> wrote: |
| 6 |
>> > The sub-roots are no longer necessary since we're installing scripts to |
| 7 |
>> > per-impl directories initially. |
| 8 |
>> > --- |
| 9 |
>> > gx86/eclass/distutils-r1.eclass | 9 +++------ |
| 10 |
>> > 1 file changed, 3 insertions(+), 6 deletions(-) |
| 11 |
>> > |
| 12 |
>> > diff --git a/gx86/eclass/distutils-r1.eclass b/gx86/eclass/distutils-r1.eclass |
| 13 |
>> > index 7102c58..5fcb55e 100644 |
| 14 |
>> > --- a/gx86/eclass/distutils-r1.eclass |
| 15 |
>> > +++ b/gx86/eclass/distutils-r1.eclass |
| 16 |
>> > @@ -476,9 +476,7 @@ distutils-r1_python_install() { |
| 17 |
>> > addpredict "$(python_get_sitedir)" |
| 18 |
>> > addpredict /usr/lib/portage/pym |
| 19 |
>> > |
| 20 |
>> > - local root=${D}/_${EPYTHON} |
| 21 |
>> > - [[ ${DISTUTILS_SINGLE_IMPL} ]] && root=${D} |
| 22 |
>> > - flags+=( --root="${root}" ) |
| 23 |
>> > + flags+=( --root="${D}" ) |
| 24 |
>> > |
| 25 |
>> > if [[ ! ${DISTUTILS_SINGLE_IMPL} ]]; then |
| 26 |
>> > local PYTHON_SCRIPTDIR |
| 27 |
>> > @@ -488,13 +486,12 @@ distutils-r1_python_install() { |
| 28 |
>> > |
| 29 |
>> > esetup.py install "${flags[@]}" "${@}" |
| 30 |
>> > |
| 31 |
>> > - if [[ -d ${root}$(python_get_sitedir)/tests ]]; then |
| 32 |
>> > + if [[ -d ${D}$(python_get_sitedir)/tests ]]; then |
| 33 |
>> > die "Package installs 'tests' package, file collisions likely." |
| 34 |
>> > fi |
| 35 |
>> > |
| 36 |
>> > if [[ ! ${DISTUTILS_SINGLE_IMPL} ]]; then |
| 37 |
>> > - _distutils-r1_wrap_scripts "${root}" |
| 38 |
>> > - multibuild_merge_root "${root}" "${D}" |
| 39 |
>> > + _distutils-r1_wrap_scripts "${D}" |
| 40 |
>> > fi |
| 41 |
>> > } |
| 42 |
>> > |
| 43 |
>> > -- |
| 44 |
>> > 1.8.3.2 |
| 45 |
>> > |
| 46 |
>> |
| 47 |
>> I believe this patch causes a race condition which causes |
| 48 |
>> java-config's ebuild to fail. |
| 49 |
>> |
| 50 |
>> running install_data |
| 51 |
>> running install_egg_info |
| 52 |
>> Writing /tmp/portage/dev-java/java-config-2.2.0/image/usr/lib64/python2.7/site-packages/java_config- |
| 53 |
>> 2.2.0-py2.7.egg-info |
| 54 |
>> error: /tmp/portage/dev-java/java-config-2.2.0/image//usr/share/java-config-2/config/: |
| 55 |
>> File exists |
| 56 |
> |
| 57 |
> I love when people hack build systems and have no idea about writing |
| 58 |
> build systems... |
| 59 |
> |
| 60 |
> What are you suggesting? Finding all the broken stuff and fixing it, or |
| 61 |
> restoring sub-roots? |
| 62 |
> |
| 63 |
|
| 64 |
This is a well-used package which will be broken if we apply this |
| 65 |
patch. I'm not sure we can reliably identify this kind of problem, and |
| 66 |
fixing individual packages would be a bit of work in any case.. |
| 67 |
Keeping the sub-roots code in place seems like the safer course of |
| 68 |
action here. |
| 69 |
|
| 70 |
Also FYI, I looked into java-config's setup.py, and the code has an |
| 71 |
unchecked os.mkdir call followed by an open() call. The former will |
| 72 |
fail for multiple python implementations even with a serialized |
| 73 |
install, and the latter is racy. |
Archives