User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.2b) Gecko/20020618 Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.2b) Gecko/20020618 Any bugzilla query will do this URL: http://bugzilla.mozilla.org/buglist.cgi?bug_status=UNCONFIRMED&bug_status=NEW&bug_status=ASSIGNED&bug_status=REOPENED&field0-0-0=product&type0-0-0=substring&value0-0-0=test&field0-0-1=component&type0-0-1=substring&value0-0-1=test&field0-0-2=short_desc&type0-0-2=substring&value0-0-2=test&field0-0-3=status_whiteboard&type0-0-3=substring&value0-0-3=test Cookies: LASTORDER=bugs.delta_ts%2C%20bugs.bug_status%2C%20bugs.priority%2C%20map_assigned_to.login_name%2C%20bugs.bug_id COLUMNLIST=opendate changeddate severity priority platform status resolution component votes target_milestone summaryfull Doesn't do this in other browsers, so I'm guessing it's cookies.. possibly the fact I sort by assigned_to, but do not have it in the column list? Reproducible: Always Steps to Reproduce:

OK, I'm not usre if this is a regression, but I think I understand what causes it. If a column was sorted on and then is dropped from the requested column list, buglist only checks to see if the column exists, not if it is being fetched. patch follows

Reject any columns from sort order if they are not in selectnames. Causes stale cookie to be deleted.

Comment on attachment 105713 [details] [diff] [review] Patch Err, why is this type of search incorrect? The fix is to modify Bugzilla/Search. At the top, it only joins to tables if they're in the requested columns list (this was changed for perf reasons). It should also consider the order param. This will mean moving 'order by' support into the .pm, rather than having the cgi tack it onto the end. ISTM that thats a sensible thing to do anyway.

Here's the SQL and error in case you need it SELECT bugs.bug_id, bugs.creation_ts, bugs.delta_ts, bugs.bug_severity, bugs.priority, bugs.rep_platform, bugs.bug_status, bugs.resolution, map_components.name, bugs.votes, bugs.target_milestone, bugs.short_desc FROM bugs, components AS map_components, products products_0, components components_0 LEFT JOIN bug_group_map ON bug_group_map.bug_id = bugs.bug_id AND bug_group_map.group_id NOT IN (9) LEFT JOIN cc ON cc.bug_id = bugs.bug_id AND cc.who = 38080 WHERE bugs.component_id = map_components.id AND bugs.product_id = products_0.id AND bugs.component_id = components_0.id AND (bugs.bug_status = 'UNCONFIRMED' OR bugs.bug_status = 'NEW' OR bugs.bug_status = 'ASSIGNED' OR bugs.bug_status = 'REOPENED') AND (INSTR(LOWER(products_0.name), 'test') OR INSTR(LOWER(components_0.name), 'test') OR INSTR(LOWER(bugs.short_desc), 'test') OR INSTR(LOWER(bugs.status_whiteboard), 'test')) AND ((bug_group_map.group_id IS NULL) OR (bugs.reporter_accessible = 1 AND bugs.reporter = 38080) OR (bugs.cclist_accessible = 1 AND cc.who IS NOT NULL) OR (bugs.assigned_to = 38080) OR (bugs.qa_contact = 38080) ) GROUP BY bugs.bug_id ORDER BY bugs.delta_ts, bugs.bug_status, bugs.priority, map_assigned_to.login_name, bugs.bug_id : Unknown table 'map_assigned_to' in order clause at globals.pl line 242.

bbaetz volunteered to do a more comprehensive patch. Reassigning to him.

Actually, I have a bit more work to do this week than I thought, so this may take me a while to get to

OK, then I'll fix this (still within buglist.cgi) by having it request the columns it needs before calling Search.

This is primarily a rearrangment of the existing code in buglist.cgi so that the order is analyzed before the search is called. The only really new code in it is the section that adds the columns in the order to the list of columns to be selected if they are already validated and are not already on the list.

Comment on attachment 105746 [details] [diff] [review] Patch v2 This is an evil, ugly hack. Eitherkeep this bug open, or file a new one assigned to me for hte proper fix. r=bbaetz to stop the breakage

a=myk

Checking in buglist.cgi; /cvsroot/mozilla/webtools/bugzilla/buglist.cgi,v <-- buglist.cgi new revision: 1.209; previous revision: 1.208 done

Still seeing this :-(

I stopped just short of making this a WORKSFORME.... This patch presumes that the earlier patch is still in place and moves the check until after all the non-cookie-based orders have been added.

Comment on attachment 105869 [details] [diff] [review] FURTHER patch This code is so hacky...

see comment 10

WRT comment 14, I agree. Someone should move this to Search.pm, see bug 179451 Checking in buglist.cgi; /cvsroot/mozilla/webtools/bugzilla/buglist.cgi,v <-- buglist.cgi new revision: 1.210; previous revision: 1.209 done