Bug 60415 – [GTK] Missing response for HTTP requests in Inspector when using the Soup cache

• Status: RESOLVED
• Resolution: FIXED
• Component: WebKitGTK

[Xan Lopez]

Try to load planet.gnome.org with epiphany+soup cache enabled. Some of the resource requests will lack a response according to the Inspector. The page is still loaded, so I guess we are calling didReceiveData but not didReceiveResponse for some reason. This might why the page takes a long time to load, WebKit might be timing out the requests. As an experiment try to disable the soup cache and load+reload planet gnome. At least in my machine it's *a lot* faster *without* the cache.

[Sergio Villar Senin]

(In reply to comment #0)
> Try to load planet.gnome.org with epiphany+soup cache enabled. Some of the resource requests will lack a response according to the Inspector. The page is still loaded, so I guess we are calling didReceiveData but not didReceiveResponse for some reason. This might why the page takes a long time to load, WebKit might be timing out the requests. As an experiment try to disable the soup cache and load+reload planet gnome. At least in my machine it's *a lot* faster *without* the cache.

I'm almost sure that we are not calling didReceiveData without a previous didReceiveResponse because we added some asserts to check that some months ago and I never got a hit on them. I noticed issues with planet.gnome.org but everytime I checked, they were caused by unreachable resources (in particular some icons). This means that Chrome for example was also taking a lot of time to load the page (until a timeout was triggered).

But if you say that disabling the cache improves the situation then it could be something else, will take a look.

[Sergio Villar Senin]

(In reply to comment #1)
> (In reply to comment #0)
> But if you say that disabling the cache improves the situation then it could be something else, will take a look.

I think I've just found what happens here. The problem is that I forgot to cache the status code of the responses, and thus, I'm not resetting it in the responses generated by the cache. Will file a bug in libsoup and set a dependency.

BTW I have also realized that we are caching all the headers and some of them (the so called hop-by-hop headers) should not as are are meaningful only for a single transport-level connection. Will file another bug in libsoup for this

[Sergio Villar Senin]

https://bugzilla.gnome.org/show_bug.cgi?id=649965

[Dominik Röttsches (drott)]

Upstream soup bug fixed.