test,fs: test watch of file and directory

[refack]

cleanup before and after test-fs-watchfile
Maybe fix: #13248
Maybe fix: #13377

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• commit message follows commit guidelines

Affected core subsystem(s)

test,fs

[refack]

Stress 10 x 100 -j 64
aix: https://ci.nodejs.org/job/node-stress-single-test/nodes=aix61-ppc64/1236/
osx: https://ci.nodejs.org/job/node-stress-single-test/1237/nodes=osx1010/

[Trott]

OS X stress test failed. (You will probably want to cancel it so it doesn't keep that host from being used for other test runs.)

I'm pretty sure this issue isn't about removing files but is instead about an OS-specific race condition in file watching. See #13248 (comment)

[refack]

I'm pretty sure this issue isn't about removing files but is instead about an OS-specific race condition in file watching. See #13248 (comment)

Ack. Stopped. But this PR is good to have anyway, no?

[Trott]

But this PR is good to have anyway, no?

We don't generally have tests clean up after themselves. I suppose exceptions can be made for tests that generate absurdly large files or that mess with file permissions. But generally, we don't do it because it's the responsibility of the test using tmpDir to clean it up before using it.

If there's some sort of subtlety with file watchers that makes cleaning up the file a Good Thing, then sure. I'm not aware of that, though, so you'll have to point me in the right direction to get me on board. :-D

[refack]

If there's some sort of subtlety with file watchers that makes cleaning up the file a Good Thing, then sure. I'm not aware of that, though, so you'll have to point me in the right direction to get me on board. :-D

On windows when there is a subdirectory in the tmpDir running common.refreshTmpDir(); throws.

Error: ENOTEMPTY: directory not empty, rmdir 'D:\code\node-cur\test\tmp' at Object.fs.rmdirSync (fs.js:856:18) at rmdirSync (D:\code\node-cur\test\common\index.js:152:10) at rimrafSync (D:\code\node-cur\test\common\index.js:122:7) at Object.exports.refreshTmpDir (D:\code\node-cur\test\common\index.js:158:3) at Object.<anonymous> (D:\code\node-cur\test\parallel\test-fs-watchfile.js:39:8)

[richardlau]

On windows when there is a subdirectory in the tmpDir running common.refreshTmpDir(); throws.

Odd, common.rmdirSync() catches ENOTEMPTY and attempts to common.rimrafSync() the contents (

node/test/common/index.js

Line 136 in fccc0bf

functionrmdirSync(p,originalEr){

).

[Trott]

Odd, common.rmdirSync() catches ENOTEMPTY and attempts to common.rimrafSync() the contents (

node/test/common/index.js

Line 136 in fccc0bf

functionrmdirSync(p,originalEr){

).

And I've never seen this Windows issue on CI. And no one has ever reported it except @refack as far as I know. I'm sure it's happening, but I suspect the cause is something specific and/or peculiar...

[refack]

And I've never seen this Windows issue on CI. And no one has ever reported it except @refack as far as I know. I'm sure it's happening, but I suspect the cause is something specific and/or peculiar...

Ya'll are right. I put a trace on, and found It's a local process (TortoiseGit) pre-checking the status of files in the working dir, that keeps some files locked.
I want to say I've seen it in the CI but it might have been a different case (RO files in #12835)

[refack]

Removed calls to unlink/rmdir but IMHO refreshTmpDir should stay, or merged with @Trott's #13252
CI: https://ci.nodejs.org/job/node-test-commit/10187/
CI: https://ci.nodejs.org/job/node-test-commit/10188/

[refack]

Found to be unnecessary.