Bug 209699 - Regression(r259034): UniqueIDBDatabas object is destroyed while it's still in use
Summary: Regression(r259034): UniqueIDBDatabas object is destroyed while it's still in...
Status: RESOLVED INVALID
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: WebKit Nightly Build
Hardware: Unspecified Unspecified
: P2 Normal
Assignee: Sihui Liu
URL:
Keywords: InRadar
Depends on:
Blocks:
 
Reported: 2020-03-28 00:07 PDT by Sihui Liu
Modified: 2020-03-29 00:04 PDT (History)
6 users (show)

See Also:

Attachments
Patch (6.35 KB, patch)
2020-03-28 00:14 PDT, Sihui Liu 		ggaren: review+
Details | Formatted Diff | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Sihui Liu 2020-03-28 00:07:58 PDT 
In r259034, we try deleting the UniqueIDBDatabase in UniqueIDBDatabase::connectionClosedFromClient, when UniqueIDBDatabase finds itself has no connections. The problem is in UniqueIDBDatabase::openDBRequestCancelled, m_pendingOpenDBRequests will be accessed after connectionClosedFromClient. To fix this issues, we can protect this pointer in UniqueIDBDatabase::openDBRequestCancelled, but this bug and 209618 have proved r259034 to be error-prone. In this case, we probably need to find a better way to delete UniqueIDBDatabase at proper timing.
Comment 1 Sihui Liu 2020-03-28 00:11:01 PDT 
rdar://problem/60951278
Comment 2 Sihui Liu 2020-03-28 00:14:28 PDT 
Created attachment 394794 [details]
Patch
Comment 3 Geoffrey Garen 2020-03-28 09:21:18 PDT 
Comment on attachment 394794 [details]
Patch

It would be nice to figure out how to use smart pointers to ensure memory safety here; smart pointers are much easier to get right.
Comment 4 Sihui Liu 2020-03-29 00:04:48 PDT 
(In reply to Geoffrey Garen from comment #3)
> Comment on attachment 394794 [details]
> Patch
> 
> It would be nice to figure out how to use smart pointers to ensure memory
> safety here; smart pointers are much easier to get right.

Yes, will try replacing the raw pointers and remove unnecessary protected pointers to make the ownership clearer. 

Since r259034 is reverted and this patch works as an alternative, I will move this patch to https://bugs.webkit.org/show_bug.cgi?id=209532 and land it from there.