Closed Bug 818855 Opened 13 years ago Closed 12 years ago

sizeof(nsLineBuffer<char>) is 4108, should be 4096?

Categories

(Core :: Networking, defect)

x86
Windows XP
defect
Not set
normal

Tracking

()

RESOLVED WONTFIX

People

(Reporter: neil, Assigned: erahm)

Details

(Whiteboard: [MemShrink:P3])

So as I understand it jemalloc ends up allocating us 8192 bytes and we therefore have 49.85% clownshoes...
It looks as if idea is that the stream should read in blocks of 4096 bytes. Is this an ideal buffer size? We would need to split this allocation from the tracking pointers, and fix NS_ReadLine not to require a trailing null.
Whiteboard: [MemShrink?] → [MemShrink]
How many of these do we have?
Whiteboard: [MemShrink] → [MemShrink:P3]
Assignee: nobody → erahm
Looked into the usage, in a simple browsing test I saw zero instances created/destroyed. It looks like the main user, nsFileInputStream::ReadLine (which creates a nsLineBuffer), is rarely used. The main locations I found were in nsCertOverrideService, nsCookieService (importing cookies), webrtc/LoadMonitor (reading system stats on Linux/Android). It is also exposed to add-ons which may use it more gratuitously. MediaResourceServer also uses an instance to parse http input (seems to be android only). There's a unicode conversion utility using nsLineBuffer<uint16_t> implementing a readline, but AFAICT it's never referenced.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.