Attempt fixing :sync

[dickmao]

See dickmao's comment in #92 regarding why :sync does not behave as
expected.

Previously, request--curl-sync would merely return as soon as the
curl process finished without regard to
#'request--curl-callback. This change attaches the semaphore to the
right place.

(Attempt) Fixes #92

[titaniumbones]

hmm, so the semaphore keeps checking in with the process about whether or not it's done? I'm a little confused still as to why async calls succeed when sync calls fail -- if there is still data to retrieve via accept-process-output shouldn't async calls also have trouble?

But then I don't understand async programmingvery well at all, esp not in elisp.

@alphapapa are you able to review this ?

[dickmao]

hmm, so the semaphore keeps checking in with the process about whether or not it's done?

No, the while-loop in request--curl-sync does that.

I'm a little confused still as to why async calls succeed when sync calls fail -- if there is still data to retrieve via accept-process-output shouldn't async calls also have trouble?

One, Async calls don't call request--curl-sync. Two, how do you define trouble? We don't make the same synchronous demands on async calls that we do :sync t calls.

[titaniumbones]

hmm, so the semaphore keeps checking in with the process about whether or not it's done?

No, the while-loop in request--curl-sync does that.

ah right. and semaphore is just another way to set finished to t, is that right?

I'm a little confused still as to why async calls succeed when sync calls fail -- if there is still data to retrieve via accept-process-output shouldn't async calls also have trouble?

One, Async calls don't call request--curl-sync. Two, how do you define trouble? We don't make the same synchronous demands on async calls that we do :sync t calls.

ok. since this is making poor wee head hurt... I will do what other people think is best. Are you able to pass all tests with yr branch?

[titaniumbones]

Is "semaphore" a standard name for this kind of process watcher?

[dickmao]

No? But "semaphore" is exactly what it is.

[dickmao]

I passed travis on the first attempt: https://travis-ci.org/tkf/emacs-request/builds/461595586

I appear not to be passing now. It a'int easy being green.

[alphapapa]

@alphapapa are you able to review this ?

I took a look and posted a few comments. I don't feel like I am familiar enough with Emacs's process-interacting code to pass judgement on it, but if @dickmao is, and if the tests pass, and if it seems to fix the bug, then great! :)

[titaniumbones]

@dickmao:

• are you feeling good about the state of this PR? if so, would you please be willing to:
• rebase on development and submit PR there? trying to stop pushing to master, esp when adding new features that have the potential to break older deployments.

[dickmao]

Done.