Conversation
… no longer exists on AWS; should be fixed in pricing model.
…/dxCompiler into for_testing_fulcrum_genomics_pr
Co-authored-by: Ted Brookings <ted@fulcrumgenomics.com>
--------- Co-authored-by: Ted Brookings <ted@fulcrumgenomics.com>
--------- Co-authored-by: Ted Brookings <ted@fulcrumgenomics.com>
* Fix typo in test invocation of -executableCreationParallelism * Maybe fix applet with dockerRegistry test failing due to change in order of independent tasks --------- Co-authored-by: Ted Brookings <ted@fulcrumgenomics.com>
…order is reversed (#519) Co-authored-by: Ted Brookings <ted@fulcrumgenomics.com>
| set -euxo pipefail | ||
|
|
||
| TEST=${1} | ||
| TEST=${1:-""} |
|
|
…/dxCompiler into for_testing_fulcrum_genomics_pr
| if (desc1.access == expected2Access && desc2.access == expected1Access) { | ||
| fail("Compiled applet order is reversed") | ||
| } | ||
|
|
||
| desc1.access shouldBe expected1Access | ||
|
|
||
| desc2.access shouldBe expected2Access | ||
|
|
There was a problem hiding this comment.
I dont get it why the access order could be wrong. The parallelized compilation is done on callable, and the network access is embedded in the callable object, so it would go with the callable and should not be affected by this new setting. @r-i-v-a
There was a problem hiding this comment.
@YuxinShi0423 I think the test is just asserting that the new setting doesn't change the order; this was mentioned in the spec attached to the PMUX ticket.
There was a problem hiding this comment.
I had introduced a bug that did reverse the order, but the resulting error message was unclear (it looked like the results were just wrong). I made the message for reversed order more clear and fixed the bug.
There was a problem hiding this comment.
@TedBrookings could you point me to the commit for that fix?
There was a problem hiding this comment.
d56c1cb Merge branch 'for_testing_fulcrum_genomics_pr' of github.com:dnanexus/dxCompiler into for_testing_fulcrum_genomics_pr
The changes were to Compiler.scala and ExecutableTree.scala. That said, this was fixing a bug that I had introduced earlier. I had been using Map keys instead of Vectors to track dependencies, so the order of mutually-independent executables had not been preserved correctly.
There was a problem hiding this comment.
Thanks! So the access config is just the flag we used to check if applet1 and 2 are in correct order, but itself wont not be swapped.
I would still prefer another unit test since this one serves two purposes now. But I am okay with it. @r-i-v-a
YuxinShi0423
left a comment
There was a problem hiding this comment.
I got no more question. Thanks! LGTM
DEVEX-2579 For testing Fulcrum Genomics PR