358208 - virtual folder's name begin with // results in "A folder with that name already exists. Please enter a different name."

Status: NEW

(Thunderbird :: Mail Window Front End, defect)

Description

[captainbou]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.0.7) Gecko/20060909 Firefox/1.5.0.7
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.0.7) Gecko/20060909 Firefox/1.5.0.7

I've tried with 3 or 4 different names like // TODO.

Reproducible: Always

Steps to Reproduce:
1. Go to File>New>Virtual Folder
2. In the name field, put "// something"
3. Click on Ok


Actual Results:  
The application freezes

Comment 1

[Magnus Melin [:mkmelin]]

Though not *exactly* the same, this sounds like a duplicate of bug 29926.

FWIW, on linux latest branch build, there is no freeze, it seems to do nothing at all (on imap, pop seems to work ok).

Comment 2

[ovidiu]

Attachment: tb3.0apre alert capture on bug358208

create saved search with // name, say ok -nothing, on second ok pops this alert

Comment 3

[ovidiu]

os: vista, test pop gmails

I confirm this behavior still goes on Tb 2.0.0.14 and TB 3.0apre (with comments)

Tb 2.0.0.14
I tried for 
new/saved search -> name: // something and for 
search messages/save ->  // something
same, hangs, need to kill app, always

Tb 3.0apre 
new/saved search -> name: // something and for 
search messages/save ->  // something
(both)
it does nothing on the first click (ok), 
-on the second [ok] alert:
"a folder with that name already exists" , where it does not exist, nor create it. Created an attachment (id=319027). Then say ok. Say cancel to the search panel. 
-or (another try or after above!) on second click [cancel], close the panel, looks like canceled,

for both cases:
1. If go to another account, seams ok, can go back to msg on the nasty acc and see them.
2. On collapse the account in folder pane [+] (for try to refresh folder tree) it hangs app, even if after step 1. above, see below a crash report

I would confirm for Tb3.0apre too, though it may have evolved in a more refined "killer". It does not do it immediately, but affects folder pane/account behavior. Note that I tried the normal +/- behavior on that account, and it goes ok, even after other saved searches, except // named ones.


fwiw here's the crash report on 3 (did also submitted that with some comments..):
Add-ons: {e2fda1a4-762b-4020-b5ad-a41df1933103}:0.6a1,{46551EC9-40F0-4e47-8E18-8E5CF550CFB8}:0.5.6,{972ce4c6-7e08-4474-a285-3208198ce6fd}:2.0
BuildID: 2008042303
CrashTime: 1209738464
InstallTime: 1208987180
ProductName: Thunderbird
StartupTime: 1209738217
Theme: classic/1.0
URL: 
UserID: 726bb6e2-5e4a-4162-8754-efe7350d5b02
Vendor: 
Version: 3.0a1pre

Comment 4

[Wayne Mery (:wsmwk)]

ovidiu 
What's the breakpad ID# for that crash? What's your opinion on this being a dupe of bug 29926?

Comment 5

[ovidiu]

how/where do I find breakpad id# ?

about bug 29926, I don't think so. First, did try on pop and saved searches, no imap. Second, tried same saved searches (as in this bug here) with only one / on name and goes fine in TB 2.0.0.14 and TB3.0apre . So I come to think is about different issues.

I also just tried on gmail-imap to create / and // named subfolders and goes ok, only that it actually names them _test and __test. Well, gmail may be a specific case here (labels vs folders..) and I don't have another imap to test with. More on the this in 29926 comments ..

Comment 6

[ovidiu]

This is dupe of bug 286523 !

But I get identical behavior in bug 312265, same procedures, same alert, same tb 2.0.0.14 + Tb 3.0apre as in comment #3 here, only using names like LPT1.

I think the correct thing will be to dupe both this and bug 286523 for bug 312265. Cause 312265 seams to cover it more extensively while this and the dupe are only for the specific case of / character. 

Note that https://bugzilla.mozilla.org/show_bug.cgi?id=286523#c8
states that virtual folders use older code than normal folders, which may lead to solution.

I'll just comment on those and add anything after dupes are set (the whole content of comment 3 + attach are applicable there too) 

Comment 7

[Wayne Mery (:wsmwk)]

(In reply to comment #5)
> how/where do I find breakpad id# ?

http://kb.mozillazine.org/Breakpad#Location_of_crash_reports

or install nightly tester tools extension - an item noted at http://wiki.mozilla.org/Thunderbird:Bug_Triage#Triage_Resources.2C_Tools_and_Hints

Comment 8

[ovidiu]

about comment #3 crash report
ID: d08805fa-1855-11dd-b93a-001cc45a2ce4

about comment #6
correction
I would dupe this for bug 312265 which has more general descr and because I reproduce the exact behavior as here c#3

as for bug 286523, I have doubts, not same test results ..

Comment 9

[timeless]

Signature	ntdll.dll@0x61636
UUID	d08805fa-1855-11dd-b93a-001cc45a2ce4
Time	2008-05-02 07:27:44-07:00
Uptime	247
Product	Thunderbird
Version	3.0a1pre
Build ID	2008042303
OS	Windows NT
OS Version	6.0.6000
CPU	x86
CPU Info	GenuineIntel family 6 model 15 stepping 2
Crash Reason	EXCEPTION_STACK_OVERFLOW
Crash Address	0x77b91636
Comments	reproducing bug 358208 see the comment there (from ovidiu), also for 2.0.0.14 both for: new/saved search -> name: // something and for search messages/save -> // something it does notthing on the first click (ok), -on the second [ok] alert: "a folder with that name already exists" , where it does not exist, nor create it. -on second click [cancel], close the panel, looks like canceled, for both cases, collapse/uncollapse account [+] for refresh hangs app
Crashing Thread
Frame 	Module 	Signature 	Source
0 	ntdll.dll 	ntdll.dll@0x61636 	
1 	ntdll.dll 	ntdll.dll@0x66863 	
2 	ntdll.dll 	ntdll.dll@0x66a0a 	
3 	ntdll.dll 	ntdll.dll@0x3cad2 	
4 	kernel32.dll 	FindFirstFileExW 	
5 	kernel32.dll 	FindFirstFileA 	
6 	msvcr80.dll 	msvcr80.dll@0x25a9a 	

ovidiu: can you try using:
http://developer.mozilla.org/en/docs/How_to_get_a_stacktrace_with_WinDbg

to get a stack trace? (!analyze -v -f)

luser: this is the shortest stack overflow i've ever seen, thoughts?

Comment 10

[(not currently active) Ted Mielczarek]

Might just be the lack of symbols for those system libraries killing the stack trace.

Comment 11

[ovidiu]

Attachment: windbg stack copy in a txt file

windbg stack copy in a txt file, in addition to crash reports

Comment 12

[ovidiu]

for comment #9
created another crash report, maybe this time..
Crash ID: bp-52bb58b5-1a30-11dd-98f4-001cc45a2ce4

for windbg, I tried, not sure if i got what you need (have never used..)
see attach in comment #11

(was that what u r lookin for?)

Comment 13

[timeless]

right idea, but this http://symbols.mozilla.org/firefox should probably be http://symbols.mozilla.org/thunderbird :) run it again w/ that included in your sympath and we should get better results.

luser: can you try to pull whatever version the reporter has into the crash-stats?

Comment 14

[ovidiu]

Attachment: second time

Comment 15

[ovidiu]

(In reply to comment #13)
> right idea, but this http://symbols.mozilla.org/firefox should probably be
> http://symbols.mozilla.org/thunderbird :) run it again w/ that included in your
> sympath and we should get better results.
> 

ok, done that and add second attach txt, though not sure if much difference (do understand that I have no idea what I'm doing here and if missing something ..)

> luser: can you try to pull whatever version the reporter has into the
> crash-stats?
> 
vista
TB: version 3.0a1pre (2008042303)
crashreporter.exe file version 1.9.0.3034 (24.04.2008)
(??)
note: when starting from windbg, no moz reporter shows

note on TB steps, to be clear:
-create saved search
-try to save as "//" name
-ok =nothing ..
-ok =alert, as pic in attach
-ok in alert
-cancel save, close [x] search panel (no ssearch saved, no change in acc folders ..)
-go to folderpane, collapse respective account (not folders or subfolders, they work ..)
- not responding/freeze/hang ..

Comment 16

[timeless]

:(, this time you didn't run to the problem point. you need to run to right before the problem point, then attach the debugger w/ the right sympath, then debug>go (to continue), trigger the problem.

if it's a crash:
!analyze -v -f (as before)

if it's a hang
debug>break
!analyze -v -hang

Comment 17

[ovidiu]

Attachment: 3rd time [and counting .. :) ]

Comment 18

[ovidiu]

(In reply to comment #17)
> Created an attachment (id=319396) [details]
> 3rd time [and counting .. :) ]
> 

please don't get it wrong, but am I the only one getting this hang? cause I always get identical results, as for tb 2 the result is even more direct. This also fully apply to bug 312265 (if I use LPT1 as name ..)

Comment 19

[ovidiu]

This is still same for 
In 3.0a1rc1 ?
Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9pre) Gecko/2008050715
Lightning/0.6a1 Thunderbird/3.0a1 ID:2008050715

though it seams that LPT1 and PRN there have changed behavior a bit bug 312265
(Also got hang with COM1 COM3 name)

Note in the comments that having "something//" had small issue too, refresh resulting in 
-appearent duplicate folders and subfolders
gigel// -original
gigel// +??
 gigel//
   gigel//
-name of folder and msf in profile gigel9d4b2d35

I'm still thinking of a way to consolidate these special character/name searches in a way, though they seam to have changed a bit in latest 3alpha rc. I'll look more into it ..

Comment 20

[WADA:World Anti-bad-Duping Agency]

To ovidiu:

Your problem is a variation of phenomena after Bug 286523(problem after problem of Bug 286523).

When virtual folder name contains "preceding slash" or "trailing slash" or "consecutive slash's", consecutive "/"s appear in mailbox: uri in virtualFolders.dat. If "/ X / Y // Z //", it becomes as follows due to Bug 286523.
> uri=mailbox://<server_name>/<parent_folder_name>// X / Y // Z //
Because "/" is delimiter of URI path, when the URI represents directory hierarchy, "/" means "go to lower directory". So "//" becomes "null directory name".
See also Bug 436032 for similar problem on usual local mail folder produced by similar cause to Bug 286523(raw "/" in mailbox: uri).
Please note that CON,PRN,LPTn cases are different issue by different cause,  although .../CON(PRN,LPTn etc.)/... case becomes similar to your consecutive "/" case(null directory name case) because MS Win won't create directory or file of CON(PRN,LPTn etc.) due to MS Win's restriction.
(i.e. CON case is similar to your "//" case, instead of that your "//" case is similar to or same as CON case, even though CON case was reported earlier.)

I attached test result(what file/directory is created on MS Win at each step) of (1) "a slash at mid", (2) "preceding slash", (3) "trailing slash" cases to Bug 286523 Comment #11. As seen in my test result, virtual folder is corrupted since initial creation due to Bug 286523. So anything wrong can occur if the corrupted virtual folder or garbage folder is accessed.
Check virtulFolders.dat content and directory/file structure after each step.

> -name of folder and msf in profile gigel9d4b2d35

As I wrote Bug 286523 Comment #10, if you rename virtual folder from "ascii only name" to "name with slash", "/" in virtual folder name is hashed as expected upon folder file renaming, and virtualFolders.dat contents is written as expected.

Comment 21

[ovidiu]

(In reply to comment #20)
> Your problem is a variation of phenomena after Bug 286523(problem after problem
> of Bug 286523).
> 
Would that make it even dupe (I was about to change same dependent..)?

(I do watch bug 286523)
And yes, LPT etc sys names already seams to be different (as in comment 19) a bit even in behavior in pre vs rc1

Comment 22

[ovidiu]

Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.1.4pre) Gecko/20090915 Thunderbird/3.0b4
Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.1.4pre) Gecko/20090922 Shredder/3.0pre
Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.9.1.4pre) Gecko/20090925 Shredder/3.0pre

latest behavior is not hang but:
-Create a saved search, choose some crit +some folders to search
-Name it //(just "//"), or COM1 or COM2 or PRN
-Say OK once, nothing happens, takes 2 sec for the ok button to get "unpressed"
-Say OK second time, alert: 
"A folder with that name already exists. Please enter a different name."

It does not exist and it's not created, but no hang.

With LPT name (NOT LPT1 !)I still get a Hang. In the first OK click, the OK button does not revert to unpressed and it hangs on that state.

Comment 23

[Wayne Mery (:wsmwk)]

confirming based on ovidiu