Mozilla 0.9 failed my PAC smoketesting while trying to get 53080 verified. Other platforms work. Nobody has ever said this works in Mac, can anyone confirm?

-->PAC bugs to Jussi

mass move, v2. qa to me.

I see the PAC file load from my webserver, but Mac fails to browse afterwards. This blocks testing of the fix for 76649, so I've moved it's kewords to here, and added some more. Needs target milestone. I will not pass this feature as QA until this is fixed.

Serge, can you take a look at these PAC issues? Thanks - Jussi-Pekka

I reproduce it with Mozilla/5.0 (Macintosh; U; PPC; en-US; rv:0.9.2+) Gecko/20010712, the browser gets the proxy.pac from the webserver,but then the browser dosen't use it.

UPDATE: Commercial, 2001-07-17-03-0.9.2 MacOS does make a direct connection, rather than failing. Did someone patch/fix this? I see it in about:cache, so I double-verify it loaded.

Where's the testcase? (Or should the bug not have the 'testcase' keyword? :-) I'm not in the mood to set up a proxy server and write a PAC file myself just to test this.

If you wanted to, you could just create any PAC file that points to a non-existen Proxy server. Since it is broken, you would find surfing would give you no errors, because the non-existent proxy reference is not used because the Mac always goes out DIRECT. http://www.packetgram.com/pktg/proxy/pac/pactest.html#smoketest These test cases need to be finalized and copied from packetgram to mozilla eventually.

We have a fix! 10 minutes with pinkerton and his mac helped work this out. See bug 92516. Summary: on the mac, we only get the IP, then we were truncating this at the first period. Oops. The dns service was then trying to resolve this invalid part of an IP, and failing, so the assignment in nsproxyautoconfig.js failed. Of course, we should have been more robust anyway. Obvious patch coming; seeking r/sr.

sr=darin

<pinkerton> bbaetz: r=pink

Approved for 0.9.3 on behalf of drivers assuming you guys open another bug to find the real problem in the dns code.

FIXED.

bbaetz: please convince CW or whatever to expand tabs per the modeline -- no big on this temporary patch, just for next time, eh? /be

I removed the tabs before checking in - or is that what you were complaining about? The modeline implies tabs, because its missing indent-tabs-mode: nil. I used xemacs, but changed it to match the rest of the file when I did my cvs diff sanity check before committing. I forgot to mention - the 'real' bug was filed as http://bugzilla.mozilla.org/show_bug.cgi?id=92516.

RELNOTE: PAC is not supported for MacOS in this release.

V: This has been working since 0.9.4, including Netscape 6.whatever (6.1) that came off that branch.