Bug 232982 – Get daemon API ready to work in the iOS simulator (starting with WebPushD)

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKit Misc.

[Brady Eidson]

Get daemon API tests working in the iOS simulator (starting with WebPushD.PermissionManagement)

[Brady Eidson]

Created attachment 443900 [details]
Patch for EWS v2

[Brady Eidson]

Created attachment 443902 [details]
Patch for EWS v2

[Brady Eidson]

Created attachment 443905 [details]
Patch for EWS v3

[Brady Eidson]

Created attachment 443906 [details]
Patch for EWS v4

[Brady Eidson]

Created attachment 443907 [details]
Patch for EWS v5

[Brady Eidson]

Created attachment 443964 [details]
Patch for EWS v6

[Brady Eidson]

Created attachment 443998 [details]
Patch for EWS v7

[Brady Eidson]

Created attachment 444004 [details]
Patch for EWS v8

[Brady Eidson]

Created attachment 444013 [details]
Patch for EWS v9

[Geoffrey Garen]

Comment on attachment 444013 [details]
Patch for EWS v9

r=me

[Brady Eidson]

Created attachment 444033 [details]
Patch for landing v1

[Brady Eidson]

Created attachment 444078 [details]
Patch for landing v2

Daemon failed to launch on the EWS bot, causing TestWebKitAPI crashes.

This patch fixes those crashes and turns them into "clean failures", but I'm still exploring why the daemon isn't launching.

[Brady Eidson]

Okay I figured out the difference between my setup and what the bots are doing.

Locally, we often symlink our built frameworks into the sim runtime.
When I did that, the tests were working great.

If I undo that and just rely on the run-api-tests script like the bots do, I get the same "daemon won't launch" failure.

Trying to figure out the difference now.

[Brady Eidson]

TL;DR

OSLaunchdJob + the simulator require us to pass a daemon path relative to the runtime root.

But run-api-tests just sets a whole bunch of Mac-relative DYLD* paths, and we don't actually *have* anything installed inside the runtime root.

Jonathan helped me brainstorm possible solutions an has volunteered to hack on run-api-tests to implement a solution that seems reasonable.

[Brady Eidson]

(In reply to Brady Eidson from comment #14)
> TL;DR
> 
> OSLaunchdJob + the simulator require us to pass a daemon path relative to
> the runtime root.
> 
> But run-api-tests just sets a whole bunch of Mac-relative DYLD* paths, and
> we don't actually *have* anything installed inside the runtime root.
> 
> Jonathan helped me brainstorm possible solutions an has volunteered to hack
> on run-api-tests to implement a solution that seems reasonable.

https://bugs.webkit.org/show_bug.cgi?id=232982 for that change.

I'll be testing changes to this patch with that one applied \o/

[Brady Eidson]

(In reply to Brady Eidson from comment #15)> 
> https://bugs.webkit.org/show_bug.cgi?id=232982 for that change.
> 
> I'll be testing changes to this patch with that one applied \o/

I, of course, meant https://bugs.webkit.org/show_bug.cgi?id=233080

[Brady Eidson]

Created attachment 444133 [details]
Patch for landing v3

[Brady Eidson]

Retitling:
        Get daemon API ready to work in the iOS simulator (starting with WebPushD)

Because I'm going to land this patch as-is - doing all the groundwork to enable the daemon tests - but still disabled until we resolve the "how to get the daemon executable visible inside the runtime root" patch

[Brady Eidson]

Created attachment 444209 [details]
Patch for landing v4

[EWS]

Committed r285799 (244243@main): <https://commits.webkit.org/244243@main>

All reviewed patches have been landed. Closing bug and clearing flags on attachment 444209 [details].

[Radar WebKit Bug Importer]

<rdar://problem/85400420>