[SPARK-10952] Only add hive to classpath if HIVE_HOME is set.

[kevincox]

Currently if it isn't set it scans /lib/* and adds every dir to the
classpath which makes the env too large and every command called
afterwords fails.

[JoshRosen]

Jenkins, this is ok to test.

[SparkQA]

Test build #43277 has finished for PR 8994 at commit 37c02b0.

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

[JoshRosen]

Not a bash expert, but should this be using something like if [ -d "$HIVE_HOME" ]; to check that it's actually a directory and exists, similar to how we check for the existence of the assembly directory in

spark/bin/spark-class

Line 51 in 5952bdb

if [ -d "$ASSEMBLY_DIR" ]; then

?

[kevincox]

I really don't know. I was thinking "If you set the variable we will try to use it" plus if isn't correct the wild card will expand to nothing anyways. The other option is to warn if HIVE_HOME is set but doesn't point to a directory. I'm not too convinced either way but the way it is now makes sense to me. Although the warning won't hurt.

[JoshRosen]

Yeah, I only commented because I'm actually somewhat wary of bash-isms and http://unix.stackexchange.com/a/32211/42294 came to mind.

[kevincox]

I think I'm good here because it is in quotes so it will always be an argument. However there is actually a problem if there is a leading - in the variable so I will update to prevent that.

[JoshRosen]

Minor question aside, this LGTM.

[kevincox]

If you would prefer the change I can do it, just let me know.

[SparkQA]

Test build #43305 has finished for PR 8994 at commit 72b88b2.

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

[JoshRosen]

It looks like these lines might be indented using using tabs instead of spaces; I'll just fix this myself on merge.

[JoshRosen]

LGTM now, so I'm going to merge this to master and branch-1.5.