CI: Travis tests timing out

[TomAugspurger]

E.g. https://travis-ci.org/pandas-dev/pandas/jobs/417983438

........................................................................ [ 95%]
.......................
No output has been received in the last 10m0s, this potentially indicates a stalled build or something wrong with the build itself.
Check the details on how to adjust your build configuration on: https://docs.travis-ci.com/user/common-build-problems/#Build-times-out-because-no-output-was-received
The build has been terminated

[TomAugspurger]

Seems to have first appeared in https://travis-ci.org/pandas-dev/pandas/jobs/418084453. It actually failed the PR build: https://travis-ci.org/pandas-dev/pandas/jobs/418011554, #22424.

Trying to reproduce locally now.

[TomAugspurger]

Unable to reproduce. I'm going to temporarily revert #22424.

[TomAugspurger]

cc @webartifex if you have any clues as to why this may be, though I don't suspect you do. This is a strange one :)

[jorisvandenbossche]

Sorry about merging the PR. I actually saw that travis was failing, but since it was a time-out, just thought it was a temporary glitch .. Should have restarted the failing build first before merging.

[jorisvandenbossche]

But indeed, a very strange one.

I actually don't see how that PR could have caused it. It only edited travis-36-doc.yaml, which is only used by the last doc build, which is actually passing. The failing build uses travis-27.yaml

[TomAugspurger]

Ah indeed I should have looked more closely at which one was failing. I'll dig deeper in the logs to see what packages changed.

[TomAugspurger]

diff --git a/b b/a
index d2ff0d5..5e1e74c 100644
--- a/b
+++ b/a
@@ -15,7 +15,7 @@ backports_abc             0.5                      py27_0
 beautifulsoup4            4.6.3                    py27_0  
 blas                      1.0                    openblas  
 blinker                   1.4                      py27_0  
-blosc                     1.14.3               hdbcaa40_0  
+blosc                     1.14.4               hdbcaa40_0  
 boto                      2.49.0                   py27_0  
 boto3                     1.7.62                   py27_1  
 botocore                  1.10.62                  py27_0  
@@ -52,7 +52,7 @@ freetype                  2.9.1                h8a8886c_0
 funcsigs                  1.0.2                    py27_0  
 functools32               3.2.3.2                  py27_1  
 futures                   3.2.0                    py27_0  
-gcsfs                     0.1.1                      py_0    conda-forge
+gcsfs                     0.1.2                      py_0    conda-forge
 glib                      2.56.1               h000015b_0  
 google-api-core           1.3.0                     <pip>
 google-auth               1.5.1                      py_0    conda-forge

[TomAugspurger]

Probably blosc

pandas/tests/io/test_packers.py::TestCompression::test_small_strings_no_warn_zlib PASSED                                                              [ 66%]
pandas/tests/io/test_packers.py::TestCompression::test_small_strings_no_warn_blosc FAILED                                                             [ 66%]
pandas/tests/io/test_packers.py::TestCompression::test_readonly_axis_blosc 

That last one is seemingly hanging.

Interestingly, I don't reproduce it when running just that one test.

[TomAugspurger]

I'm going to temporarily pin blosc (the C library) to 1.14.3 till we can sort this out.

[TomAugspurger]

Should be fixed by #22429. Opened #22430 for removing the blosc pin.