Closed
Bug 139292
Opened 24 years ago
Closed 24 years ago
NSS_NoDB_Init behavior change in NSS 3.4
Categories
(NSS :: Libraries, defect, P1)
Tracking
(Not tracked)
RESOLVED
FIXED
3.6
People
(Reporter: julien.pierre, Assigned: rrelyea)
Details
Attachments
(1 file, 1 obsolete file)
|
518 bytes,
patch
|
wtc
:
review+
|
Details | Diff | Splinter Review |
This function takes a pathname (configdir) as an argument. In NSS 3.3, it
doesn't appear that the configdir has much relevance. There are obviously no
cert or key DBs loaded; but the secmod.db if present is also not loaded.
In NSS 3.4, this function looks at the configdir and does load secmod.db .
We need to clarify the behavior for this API and decide which one is the correct
thing to do.
Comment 1•24 years ago
|
||
Changed the QA contact to Bishakha.
QA Contact: sonja.mirtitsch → bishakhabanerjee
Comment 2•24 years ago
|
||
Assigned the bug to Bob.
Assignee: wtc → relyea
Priority: -- → P1
Target Milestone: --- → 3.6
| Assignee | ||
Comment 3•24 years ago
|
||
The 3.3 behavior is more correct. Applications use NSS_NoDB_Init() so that it
can use NSS encryption/hashing primitives without pulling in the whole API.
bob
| Reporter | ||
Comment 4•24 years ago
|
||
Bob,
What is the purpose of the configdir argument in that function if we never use
it ? Should the function take no arguments instead ? I realize we can't change
the prototype now, but it seems odd that it had this argument that was never used.
NSS 3.4 allows the configdir parameter to be NULL, and then everything behaves
like in 3.3 . The NSS 3.4 behavior seems logical, except that it isn't
backwards-compatible with 3.3 .
Comment 5•24 years ago
|
||
Julien, I believe that at the time that the NSS_NoDB_Init funciton was
created, the only other NSS init functions were NSS_Init and
NSS_InitReadWrite, both of which took a directory name argument.
The other forms of NSS_Init had not yet been invented. It was deemed
desirable for all 3 forms of NSS_Init to have identical signatures,
so that one could create a pointer to an NSS init function that could
point to any of them. That is why NSS_NoDB_Init takes a directory
argument that it originally did not use.
I agree with bob that NoDB_Init should not open/create/modify secmod.db
any more than it opens/creates/modifies the cert and key dbs.
| Reporter | ||
Comment 6•24 years ago
|
||
| Assignee | ||
Comment 7•24 years ago
|
||
Better to change SECMOD_DB to "". The existing patch would still open secmod.db
if it happened to be in the current directory. (or "/", I'm I'd have to look at
the code).
bob
| Reporter | ||
Comment 8•24 years ago
|
||
Bob, I think both secmod and configdir should be "" . We don't want configdir
passed down to anywhere to make sure it never gets used, just like in 3.3 in
before. And we also don't want secmod.db to be opened in the current directory.
Attachment #87152 -
Attachment is obsolete: true
| Reporter | ||
Comment 9•24 years ago
|
||
Is this simple patch OK ?
| Reporter | ||
Comment 10•24 years ago
|
||
Wan-Teh, Bob, should I check this one in for 3.6 ?
Comment 11•24 years ago
|
||
Comment on attachment 87166 [details] [diff] [review]
new patch
r=wtc. Please check it in.
Attachment #87166 -
Flags: review+
| Reporter | ||
Comment 12•24 years ago
|
||
Checking in nssinit.c;
/cvsroot/mozilla/security/nss/lib/nss/nssinit.c,v <-- nssinit.c
new revision: 1.54; previous revision: 1.53
done
Status: NEW → RESOLVED
Closed: 24 years ago
Resolution: --- → FIXED
You need to log in
before you can comment on or make changes to this bug.
Description
•