Closed UltCombo closed 8 years ago
Actually if it cleans up the code lets go ahead and remove regexp support, I've never seen anyone use it. No need to keep that feature - it isn't even documented well afaik
Something odd is that Mocha was seemingly "cascading" certain test fails -- e.g. in my first commit above, the "glob-stream create() should handle RegExps as negative matchers" test shows up as a fail, but it passes when run in isolation.
I've removed RegExp support, but no change in the failed tests. I wonder if it is an issue with glob's ignore
option, or if I'm just too sleepy to see an obvious error.
Oh -- I'm forgetting to unrelative
the negative globs, doh.
Looks like there is still an issue with unrelative logic.
Oh wait, the unrelative logic is fine -- node-glob's ignore option seems to be screwing it up independent of using relative or absolute paths.
"should return a file name stream with dotfiles negated" is failing due to a node-glob issue.
I thought the error could be due to passing absolute paths (which use \
as dir separator in Windows) to the ignore
option (according to the docs, node-glob always interprets \
as escape sequences).
Minimatch@2 does treat \
as a valid dir separator, but node-glob currently does not.
The test is still failing on Travis (Linux), so there must be yet another error...
Any progress on this?
@contra there are some blocking issues:
@contra If you don't mind the breaking change regarding the dot
behavior, we can skip those tests temporarily and I can try fixing the remaining logic again.
@UltCombo Seems like the resolution from the glob issue was the ball is in your court. Do you need any help with this?
https://github.com/isaacs/node-glob/pull/166#issuecomment-76292453
@contra Well yes, I have near to no experience with the node-glob code base, and a good part of it seems nearly unreadable to me.
I nearly have a working PR for this, there's one test broken: 'should return a file name stream with dotfiles negated' and it's timing out at 2000ms.
Any thoughts on what that might be, for some reason the stream is not emitting anything. https://github.com/thirdcreed/glob-stream/blob/master/index.js
https://github.com/isaacs/node-glob/issues/232 This is a block for me. I think there's some bigger problem involving globs that end with "/**", a completely different matcher is used for those cases. There's some kind of race condition. It emits before the on end event is registered.
I'm going to wait until issacs has time to answer it, I guess this above my head. Apparently the race condition thing is understood, It looks like its on the user to guarantee their events. Unless issacs has some insight on why their would be a race condition on the this one end event, in this one edge case. I’m out of ideas. I've been hacking on my own node-glob, but I can't seem to get that one fucking test to run.
It looks like @UltCombo's on this again anyway.
There are 23 failing tests (on both current master and @thirdcreed's branch) when testing on Windows. This does not make it easy to work on a fix and I'm relatively short on free time.
I guess I can give this another shot on my Linux or Mac box soon.
@contra May I suggest adding AppVeyor (Windows CI)? Here's a sample appveyor.yml just in case.
@UltCombo We use appveyor on some other stuff, feel free to add it if you need it
@contra Do you have an appveyor config of your choice to add to this repository? Or should I just create one using the same Node.js versions as this repo's Travis config?
Something odd is that Mocha was seemingly "cascading" certain test fails
if there are multiple tests with the same description, mocha will run them all, even if one has .only
on it. I remember this driving me crazy a few times before @doowb noticed what was happening and pointed it out. again, not sure if it's related. but if not, maybe it will save someone else from going crazy lol.
Closing in favor of #71 which is passing all the tests. Thanks for putting in a lot of the work here @UltCombo
Oh, this was still on my TODO list, but kept falling off the radar.
No problem, and indeed, #71 seems more suitable and up-to-date. I'll comment there if I find anything odd.
Quite a few tests are broken on my machine, not sure if I'm missing something or the
ignore
option is buggy in regard to thedot
option and absolute paths (on Windows).Fixes #24