Closed Bug 179238 Opened 22 years ago Closed 22 years ago

Searching by commenter (longdesc) is ignored if you do "exact" or "contains"


(Bugzilla :: Query/Bug List, defect)

Not set



Bugzilla 2.18


(Reporter: gerv, Assigned: bbaetz)





(1 file)

In bug 179176, Andrew Schultz says:

"searching by commenter (longdesc) seems to be ignored if you do "exact" or
"contains".  "matches regexp" seems to work."

This bug is to track that issue. - just for future
reference, please file bugs for issues rather than adding comments. Thanks :-)

Blocks: 179176
right, but I would assume that you were a "commenter" on all those bugs. 
removing the email criteria still returns 28 bugs.

here's the type of search I was using.   Reopened bugs that were changed
yesterday or today.  28 of them, but I haven't commented on any of them.
Gerv, if you want new bugs instead of comments, I suggest that
you don't have this banner text after upgrades :

Note: Bugzilla has been upgraded. Regressions are being tracked in meta-bug
179196. If you think you have found a regression, report it in that bug. 

A little unfair on Andrew I believe ;)
Gerv, reference comment #0, for the record, the banner at the top says "If you
think you have found a regression, report it in that bug."  Andrew just did as
he was told :)
Ya, I was contemplating filing a new bug, but figured I would follow
directions... :)
And this search

returns 435 bugs... 
Yep, happens for me too, email -is- returns
all open bugs, instead of ones I commented on. when I go back to
the search page, the email field has been wiped out.
CC bbaetz - this is the bug we talked about
<bbaetz> can you file a blocker bug, blocking 'bmo-regressions'

Raising severity. I use the address fields in almost half my searches
Severity: normal → blocker
note that "regexp" searches still work and is a viable workaround for "is" and
"contains" (probably a little slower, but...)
rkaa gave me sql last night from the buglist

The problem is that we join to the longdescs table, but we don't add a
constraint to the where clause which sort of defeats the purpose. I'll look into
this in a bit.
bugzilla, not m.o

Joel, this is your regression:

push(@supptables, "LEFT JOIN longdescs $table ON bugs.bug_id = $table.bug_id AND
$table.who IN($list)");
                push(@clist, "$table.who",'isnotnull');

@clist only has two items, so we bail out.

I'm going to drop the isnotnull branch you added, and just push to wherepart
manually. Or I could add a dummy item as the third element, but I think that the
first is cleaner.
Assignee: myk → bbaetz
Component: Server Operations → Query/Bug List
Product: → Bugzilla
QA Contact: myk → mattyt
Target Milestone: --- → Bugzilla 2.18
Version: other → 2.17.1
Attached patch fixSplinter Review
I also have to push something to @ctype so that we don't trigger the error a
few lines down
I'm getting a lot of msqld-watcher mails at the moment. This could be the cause
- the below query probably takes a lot of work.

7573   bugs   localhost   shadow_bugs   Query     184     Sorting result  
SELECT bugs.bug_id, bugs.bug_severity, bugs.priority, bugs.rep_platform,
       map_assigned_to.login_name, bugs.bug_status, bugs.resolution,
 FROM bugs, profiles AS map_assigned_to 
   LEFT JOIN longdescs longdescs_ ON bugs.bug_id = longdescs_.bug_id 
         AND longdescs_.who IN(73509) 
   LEFT JOIN bug_group_map  ON bug_group_map.bug_id = bugs.bug_id  
   LEFT JOIN cc ON cc.bug_id = bugs.bug_id 
         AND cc.who = 73509  
WHERE bugs.assigned_to = map_assigned_to.userid 
  AND 1 = 1
  AND ((bug_group_map.group_id IS NULL)    
       OR (bugs.reporter_accessible = 1 AND bugs.reporter = 73509)     
       OR (bugs.cclist_accessible = 1 AND cc.who IS NOT NULL)     
       OR (bugs.assigned_to = 73509) 
       OR (bugs.qa_contact = 73509) ) 
GROUP BY bugs.bug_id 
ORDER BY bugs.bug_id   

gerv: yeah, this degrades to a whole-db search if you don't ahve any extra
conditions (not that a query for 'all bugs I've commented on, ever' if going to
be fast, mind you)
Attachment #105724 - Flags: review?
Comment on attachment 105724 [details] [diff] [review]

This works as expected.  Only one review needed, so cancelling the second. 
r=myk, a=myk
Attachment #105724 - Flags: review?
Attachment #105724 - Flags: review+
Closed: 22 years ago
Resolution: --- → FIXED
... which means fied in cvs, not bmo, although myk was going to fix this one
manually RSN, mainly becuase of the perf hit when effectivl;y trying to retreive
all bugs.
'mattyt' != our mattyt, sorry
'mattyt' != our mattyt, sorry
QA Contact: mattyt → matty
QA Contact: matty_is_a_geek → default-qa
You need to log in before you can comment on or make changes to this bug.