Commit 293ddb4d authored by Julien Muchembled's avatar Julien Muchembled

python3: switch to 3.7

parent ac909fbc
Pipeline #9236 failed with stage
...@@ -20,7 +20,7 @@ parts = ...@@ -20,7 +20,7 @@ parts =
python3 python3
[python3] [python3]
<= python3.6 <= python3.7
[python3-common] [python3-common]
recipe = slapos.recipe.cmmi recipe = slapos.recipe.cmmi
...@@ -61,12 +61,7 @@ md5sum = 986078f11b39074be22a199e56491d98 ...@@ -61,12 +61,7 @@ md5sum = 986078f11b39074be22a199e56491d98
[python3.7] [python3.7]
<= python3-common <= python3-common
version = 3.7 version = 3.7
package_version = 3.7.2 package_version = 3.7.7
md5sum = df6ec36011808205beda239c72f947cb md5sum = 172c650156f7bea68ce31b2fd01fa766
patch-options = patch-options =
patches = patches =
environment =
PATH=${xz-utils:location}/bin:%(PATH)s
CPPFLAGS=-I${zlib:location}/include -I${xz-utils:location}/include -I${readline:location}/include -I${libexpat:location}/include -I${libffi:location}/include -I${ncurses:location}/include -I${ncurses:location}/include/ncursesw -I${bzip2:location}/include -I${gdbm:location}/include -I${openssl:location}/include -I${sqlite3:location}/include -I${gettext:location}/include
LDFLAGS=-L${zlib:location}/lib -L${xz-utils:location}/lib -L${readline:location}/lib -L${libexpat:location}/lib -L${libffi:location}/lib -L${ncurses:location}/lib -L${bzip2:location}/lib -L${gdbm:location}/lib -L${openssl:location}/lib -L${sqlite3:location}/lib -Wl,-rpath=${zlib:location}/lib -Wl,-rpath=${xz-utils:location}/lib -Wl,-rpath=${readline:location}/lib -Wl,-rpath=${libexpat:location}/lib -Wl,-rpath=${libffi:location}/lib -Wl,-rpath=${ncurses:location}/lib -Wl,-rpath=${bzip2:location}/lib -Wl,-rpath=${gdbm:location}/lib -Wl,-rpath=${openssl:location}/lib -Wl,-rpath=${sqlite3:location}/lib -L${gettext:location}/lib -Wl,-rpath=${gettext:location}/lib -Wl,-rpath=${file:location}/lib
# LD_LIBRARY_PATH=${libpng:location}/lib
  • It seems that since this we have test failures in slapos.rebootstrap and slapos.recipe.template

    I believe this is because of a problem in buildout that was fixed in https://github.com/buildout/buildout/commit/978a6af9e78db44b9db25cb08c7f9d98c24bbc03 but that patch does not apply in our fork

  • mentioned in merge request erp5!1126 (merged)

    Toggle commit list
  • Still working on this. 2 days ago, I found issues about the install command (buildout install [parts]...): not important because maybe nobody uses it but one of them is that we have extra failures (in buildout.txt) when running the test suite with Python 3. But I've just found that the py2/py3 issue would not be in the install command...

    We have several problems about the buildout test suite. With Python 2, it does not run entirely (which explains why we have extra failures with Python 3) and it's like that since November 2018. Compare the following consecutive test results:

    Which is strange in 2 ways:

    • The only difference in the code between these 2 results is a commit in the ERP5 SR, which is unrelated.
    • There are little difference in the duration (539s vs 462s). Today, the difference between Py2 (89 tests, 364s) and Py3 (376 tests, 1020s) is significantly greater.
    Edited by Julien Muchembled
  • Looks like a regression in manuel. At that time, we didn't pin The buildout test suite does not pin eggs and there were new versions at that time (14/15 Nov 2018).

    Edited by Julien Muchembled
    • 1.10.0 (2018-11-14)

      • Fix DeprecationWarning about ‘U’ mode under Python 3.
      • Drop Python 2.7 and 3.3 support. Add testing and support for Python 3.6 and 3.7.

    Pff, that's crazy. 2.7 is a typo. They meant 2.6

    The issue is fixed upstream but not released. See https://github.com/benji-york/manuel/pull/21

    Edited by Julien Muchembled
  • I have fixed the way we run tests. The only remaining difference between Py2 & Py3 is that the latter is much slower.

Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment