Bug 179101 – [GTK] Problem with SquirrelMail

• Component: WebKitGTK
• Status: NEW

Login to SquirrelMail okay, but clicking menu items fails with a request for login.

Using Xubuntu 17.10 with latest version of Web (epiphany-browser) in Ubuntu repository.

WebKit          libwebkit2gtk-4.0-37
SquirrelMail    version 1.4.22

Works correctly with uzbl (WebKit+) web browser.
Also fails with Qupzilla (WebKit2 QT) web browser.

Fails in MiniBrowser (WebKit2) but works as expected in Surf-2 (WebKit2).

Surf-2 appears to use curl to download from the server and MiniBrowser lets WebKit download through libsoup. Maybe curl manages the cookies correctly?

[Michael Catanzaro]

Could possibly be https://bugzilla.gnome.org/show_bug.cgi?id=792130. Please test with libsoup 2.62.1, where I loosened the cookie policy a bit.

Looking at what is currently installed: libsoup-gnome2.4-1 libsoup2.4-1

These are the latest versions in Ubuntu repository. Newer version of libsoup not convenient. Surf-2 works as expected. The problem does not *appear* to be related to this.

[Michael Catanzaro]

(In reply to bugzilla-ok from comment #4)
> Looking at what is currently installed: libsoup-gnome2.4-1 libsoup2.4-1

That's the API version. You'll need to check the package version.

Unfortunately libsoup does control cookie policy, so to test if this is fixed you're going to need libsoup 2.62.1. Looks like Ubuntu 17.10 has libsoup 2.60.1, and I highly doubt that will be updated. But Ubuntu 18.04 has 2.62.1. Rumor has it the release date is tomorrow. Or you can build it yourself....

There's one more thing we can check: try changing the cookie policy to "always accept" in Epiphany's preferences dialog. If that fixes it, then there's a good chance the libsoup update will resolve the problem. If not, then the libsoup update will definitely not resolve the problem.

Good suggestion. That's much easier to test!

Same problem with "always accept."

[Michael Catanzaro]

OK, then we know this is not fixed in libsoup after all.

True! It's probably something simple, but it's a puzzle.

Problem persists. Any ideas what to try next?

[Michael Catanzaro]

Nope, though I'm suspecting some problem with cookies.

Makes sense. Epiphany and MiniBrowser can't read cookies saved by SquirrelMail or perhaps save them incorrectly. Surf-2 works fine, so maybe it handles cookies itself rather than relying on WebKit. WebKit1 worked fine too. Presumably something changed moving from WebKit1 to WebKit2?

Would it be helpful to continue with this bug?

[Michael Catanzaro]

(In reply to bugzilla-ok from comment #11)
> Makes sense. Epiphany and MiniBrowser can't read cookies saved by
> SquirrelMail or perhaps save them incorrectly.

Are you sure? I said I suspected a problem, that doesn't mean I'm right. Check in the Storage tab of the web inspector to see what it's doing. Make sure cookie policy is set to always save. (And remember cookies are not persistent in MiniBrowser, unless you have the latest trunk with https://trac.webkit.org/changeset/231637/webkit and are setting the right options.)

> Surf-2 works fine, so maybe
> it handles cookies itself rather than relying on WebKit. WebKit1 worked fine
> too. Presumably something changed moving from WebKit1 to WebKit2?

I dunno what Surf does. WebKit1 was such a long time ago that comparisons are not going to be helpful.

> Would it be helpful to continue with this bug?

Yes, so far all signs point to a WebKit bug, so please leave it open.

[Adrian Perez]

(In reply to Michael Catanzaro from comment #12)
> (In reply to bugzilla-ok from comment #11)
> > Makes sense. Epiphany and MiniBrowser can't read cookies saved by
> > SquirrelMail or perhaps save them incorrectly.
> 
> Are you sure? I said I suspected a problem, that doesn't mean I'm right.
> Check in the Storage tab of the web inspector to see what it's doing. Make
> sure cookie policy is set to always save. (And remember cookies are not
> persistent in MiniBrowser, unless you have the latest trunk with
> https://trac.webkit.org/changeset/231637/webkit and are setting the right
> options.)
> 
> > Surf-2 works fine, so maybe
> > it handles cookies itself rather than relying on WebKit. WebKit1 worked fine
> > too. Presumably something changed moving from WebKit1 to WebKit2?
> 
> I dunno what Surf does. WebKit1 was such a long time ago that comparisons
> are not going to be helpful.

Surf 2.x uses WebKit2, by the way. It uses the WebKitCookieManager
with the cookies as plain text (WEBKIT_COOKIE_PERSISTENT_STORAGE_TEXT).
See here:  https://git.suckless.org/surf/tree/surf.c#n1121

> > Would it be helpful to continue with this bug?
> 
> Yes, so far all signs point to a WebKit bug, so please leave it open.

>Surf 2.x uses WebKit2, by the way.

And Epiphany uses WebKit2-GTK? The two are substantially different?

>It uses the WebKitCookieManager with the cookies as plain text (WEBKIT_COOKIE_PERSISTENT_STORAGE_TEXT).

Epiphany and MiniBrowser handle cookies differently? Maybe SquirrelMail cookies fail if they aren't handled as plain text?

[Michael Catanzaro]

(In reply to bugzilla-ok from comment #14)
> >Surf 2.x uses WebKit2, by the way.
> 
> And Epiphany uses WebKit2-GTK? The two are substantially different?
> 
> >It uses the WebKitCookieManager with the cookies as plain text (WEBKIT_COOKIE_PERSISTENT_STORAGE_TEXT).
> 
> Epiphany and MiniBrowser handle cookies differently?

Epiphany uses an SQLite database. MiniBrowser just stores them in memory until you close the browser, unless you use the command line flags that Carlos Garcia added last week, in which case you can choose whether it uses SQLite or a text file.

> Maybe SquirrelMail
> cookies fail if they aren't handled as plain text?

I suppose it might be possible. Should be straightforward to test by building MiniBrowser from WebKit trunk. Look over https://trac.webkit.org/changeset/231637/webkit to review how to switch from SQLite to text database format by changing the filename of the cookie storage file.

It sounds like when you say plain text cookies, you are talking about how they are stored rather than what is exchanged. Problems with storage might be a problem, but not having trouble anywhere but SquirrelMail. Also, MiniBrowser fails with cookies in memory, so method of saving to disk shouldn't matter. The problem would seem to be something else to do with cookies. Or perhaps JavaScript?

[Adrian Perez]

(In reply to bugzilla-ok from comment #16)
> It sounds like when you say plain text cookies, you are talking about how
> they are stored rather than what is exchanged. Problems with storage might
> be a problem, but not having trouble anywhere but SquirrelMail. Also,
> MiniBrowser fails with cookies in memory, so method of saving to disk
> shouldn't matter. The problem would seem to be something else to do with
> cookies. Or perhaps JavaScript?

Yes, of course I was referring to the format in which cookies are
stored on disk. As you cleverly wrote, the fact that MiniBrowser has
the issue with cookies in memory should make us think that the problem
is elsewhere and not in the handling of on-disk cookie persistence.

[Michael Catanzaro]

Does squirrelmail use any redirects? Could it be bug #184948?

Created attachment 340527 [details]
SquirrelMail source

Yes, SquirrelMail does redirect. I have attached the source. . . .