Uh oh!
There was an error while loading. Please reload this page.
[SPARK-20727] Skip tests that use Hadoop utils on CRAN Windows - #17966
[SPARK-20727] Skip tests that use Hadoop utils on CRAN Windows#17966shivaram wants to merge 2 commits into
Conversation
This change skips tests that use the Hadoop libraries while running on CRAN check with Windows as the operating system. This is to handle cases where the Hadoop winutils binaries are missing on the target system. The skipped tests consist of 1. Tests that save, load a model in MLlib 2. Tests that save, load CSV, JSON and Parquet files in SQL 3. Hive tests
shivaram
commented
May 12, 2017
cc @felixcheung FWIW it might easier to view the diff by adding |
vanzin
commented
May 12, 2017
This doesn't seem to really have anything to do with the errors in SPARK-20666. As I pointed out, they do not only happen on Windows. |
shivaram
commented
May 12, 2017
Sorry @vanzin I got the wrong JIRA number. Fixing it now |
shivaram
commented
May 12, 2017
This is SPARK-20727 - I just happened to have the other JIRA also open and pasted it incorrectly |
SparkQA
commented
May 12, 2017
Test build #76876 has finished for PR 17966 at commit
|
shivaram
commented
May 12, 2017
Actually thinking more about this, I think we should be checking for availability of |
shivaram
commented
May 12, 2017
@HyukjinKwon Do we know why things sometime queue for a long time on AppVeyor ? Like this PR has been queued for around 5 hours right now. |
HyukjinKwon
commented
May 13, 2017
Thank you for cc'ing me. I think primarily it is because single AppVeyor account is shared across several Apache projects but the number of concurrent jobs is single up to my knowledge. So, it is a kind of a global queue. I am monitoring ... https://ci.appveyor.com/project/ApacheSoftwareFoundation/spark |
HyukjinKwon
commented
May 13, 2017
Just FYI, closing and opening a PR is a workaround to re-trigger the build in AppVeyor as (I assume) we all don't currently have the permission via AppVeyor Web UI. |
felixcheung
commented
May 13, 2017
via email
Probably - but how to check for Hadoop? See if HADOOP_HOME is set?
We don't need to set that on *nix though, I think |
So I'd propose this |
felixcheung
commented
May 19, 2017
@shivaram have you got a chance to work on this again? |
shivaram
commented
May 20, 2017
Sorry I've been out traveling -- I'll try to update this by tonight |
shivaram
commented
May 20, 2017
@felixcheung I made the change - I'm right now going to test this in my Windows VM. Will update this PR with the results |
SparkQA
commented
May 20, 2017
Test build #77114 has finished for PR 17966 at commit
|
shivaram
commented
May 22, 2017
@felixcheung Unfortunately I'm out traveling and haven't been able to do the windows tests yet -- Would you have a chance to do that ? Also what are your thoughts on merging this while we test given the upcoming RC cut ? |
felixcheung
commented
May 22, 2017
testing now - could you submit to https://win-builder.r-project.org? |
## What changes were proposed in this pull request? This change skips tests that use the Hadoop libraries while running on CRAN check with Windows as the operating system. This is to handle cases where the Hadoop winutils binaries are missing on the target system. The skipped tests consist of 1. Tests that save, load a model in MLlib 2. Tests that save, load CSV, JSON and Parquet files in SQL 3. Hive tests ## How was this patch tested? Tested by running on a local windows VM with HADOOP_HOME unset. Also testing with https://win-builder.r-project.org Author: Shivaram Venkataraman <shivaram@cs.berkeley.edu> Closes#17966 from shivaram/sparkr-windows-cran. (cherry picked from commit d06610f) Signed-off-by: Felix Cheung <felixcheung@apache.org>
felixcheung
commented
May 23, 2017
merged to master/2.2 I think we should still check win-builder. Also it's a bit hard to tell if the skipped tests are skipped - might want to follow up with a trace |
shivaram
commented
May 23, 2017
Thanks I'll try to kick off the winbuilder build soon (i'm out of town till tomorrow). One more thing we might need to fix is that winbuilder has a 10 or 20 minute time limit for tests (not sure if the CRAN check has a similar limit ?) |
## What changes were proposed in this pull request? This change skips tests that use the Hadoop libraries while running on CRAN check with Windows as the operating system. This is to handle cases where the Hadoop winutils binaries are missing on the target system. The skipped tests consist of 1. Tests that save, load a model in MLlib 2. Tests that save, load CSV, JSON and Parquet files in SQL 3. Hive tests ## How was this patch tested? Tested by running on a local windows VM with HADOOP_HOME unset. Also testing with https://win-builder.r-project.org Author: Shivaram Venkataraman <shivaram@cs.berkeley.edu> Closesapache#17966 from shivaram/sparkr-windows-cran.
What changes were proposed in this pull request?
This change skips tests that use the Hadoop libraries while running
on CRAN check with Windows as the operating system. This is to handle
cases where the Hadoop winutils binaries are missing on the target
system. The skipped tests consist of
How was this patch tested?
Tested by running on a local windows VM with HADOOP_HOME unset. Also testing with https://win-builder.r-project.org