[SparkR][Test]:remove unnecessary suppressWarnings

[wangmiao1981]

What changes were proposed in this pull request?

In test_mllib.R, there are two unnecessary suppressWarnings. This PR just removes them.

How was this patch tested?

Existing unit tests.

[SparkQA]

Test build #67827 has finished for PR 15697 at commit a292ae8.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[wangmiao1981]

retest it please

[wangmiao1981]

windows failure seems unrelated

[HyukjinKwon]

Build started: [SparkR] ALL
Diff: master...spark-test:61EB4045-4845-4B2D-9AE9-6F0CC42B7B6E

[HyukjinKwon]

Hm, yes it seems unrelated. I will look into this deeper.

[HyukjinKwon]

It seems R 3.3.2 is released but not for Windows yet (in see - https://cloud.r-project.org/bin/windows/base/old). So, it retrieves the latest version as 3.3.2. So it tries to download the older version 3.3.1 which does not exist. I guess this should be fine soon but if it is not going to release 3.3.2 for Windows, I will make the R version down to 3.3.0 for AppVeyor.

[wangmiao1981]

@HyukjinKwon Thanks for your time!

[HyukjinKwon]

Build started: [SparkR] ALL
Diff: master...spark-test:846E764A-967B-4CB2-A78C-968D86EDFEF2

I am pretty sure it is fine now :) it seems 3.3.2 was just released.

[wangmiao1981]

retest this please

[SparkQA]

Test build #67909 has finished for PR 15697 at commit a292ae8.

• This patch fails SparkR unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[wangmiao1981]

retest this please

[SparkQA]

Test build #67920 has finished for PR 15697 at commit a292ae8.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[wangmiao1981]

@HyukjinKwon It seems that windows test still fails with the same error

[HyukjinKwon]

Oh, AppVeyor test is not triggered again with the Jebkins one.
Or, do you mean you launched another build via another account and manually tested it?

[wangmiao1981]

No, I just launched Jenkins test. I don't know how to manually trigger the AppVeyor test.

[HyukjinKwon]

Ah, yes actually that's a problem now unfortunately. It seems even comitters can't retrigger. Actually, that's why I ran this via another account and left the comment above. I am trying to find a better way and will probably open a INFRA jira.

[HyukjinKwon]

Hey @wangmiao1981, I just found another (a bit ugly) workaround to re-trigger. It seems it starts to rebuild after closing and re-opening. I just tested in spark-test#11. If you are worried of the failure mark, maybe you could try.

[felixcheung]

@HyukjinKwon the time window when download of an older version of R for windows could fail seem concerning. Is there a way to address that?

[HyukjinKwon]

For the specific failure here, it was due to not syncing the latest version between the releases for Winodws and the others (when R 3.3.2 was released). As we use a fixed version, R 3.3.1 now (not the latest but old one), we wouldn't meet the same case for the same reason in the future (I described this problem in more details in #15709).

However, I guess you meant the concern of re-triggering the spurious failures in general. For re-triggering tests, I am still looking into this deeper. Will make a INFRA JIRA as soon as I figure out the way committers can re-trigger. Current ways I have figured out so far as a (ugly) workaround are,

• Use another AppVeyor account
• Closing and then re-opening to launch another build.

[wangmiao1981]

Now, I try to close and reopen it to trigger another windows test.

[wangmiao1981]

reopen it now.

[felixcheung]

right, I think I see two problems, both of those you have looked into (thanks)
first, if we can't rely on R releases for multiple platform to be push at roughly the same time, that's problematic if there is no other way to find the latest version of that platform. Since we are fixed on the lowest version on Jenkins, I see one of the benefits with appveyor is that we could pick up the latest version (we have been broken by both, more than once)

second, re-triggering is spotty. Is that perhaps an appveyor issue?

In any case we could likely follow up in separate JIRAs.

[HyukjinKwon]

Yes, it is probably an issue because it seems currently no one except who has the AppVeyor account called ApacheSoftwareFoundation can re-trigger. I should look into this further though.. Will update you.

[wangmiao1981]

@felixcheung Any comments?

[felixcheung]

merged to master.