27416 - nsIFile performance needs evaluation

Status: RESOLVED INCOMPLETE

(Core :: XPCOM, defect)

Description

[Simon Fraser [no longer active]]

I suspect that we lost a number of performance enhancements in the move from 
nsFileSpec to nsIFile. We need to evaluate nsIFile for performance, particularly 
on Mac.

Comment 1

[Doug Turner (:dougt)]

*** Bug 27079 has been marked as a duplicate of this bug. ***

Comment 2

[Doug Turner (:dougt)]

later.

Comment 3

[Simon Fraser [no longer active]]

Why is this latered? We have serious startup performance problems, some of which 
may be relate to the transition to nsIFile from nsFileSpec. We have done *no* 
performance evaluation of the various nsIFile implementations on each platform.

Reopening.

Comment 4

[Doug Turner (:dougt)]

It was latered based on my evaluation of my task load.  

What "performance enhancments" do you suspect were lost?

Comment 5

[Simon Fraser [no longer active]]

Well, in the Mac implementation, I suspect we are calling ResolveAlias and 
PBGetCatInfo multiple times per file, for example. We did suffer a significant 
slowdown in startup performance around the time that nsIFile landed, so I do 
think that it is important to look at this.

If you like, file separate bugs for each platform, and then platform gurus can 
look at them.

Comment 6

[Doug Turner (:dougt)]

I think that this is a mac-only problem as on windows I have made sure that 
resolving shortcuts are only called when a file does not exists and you call 
stat.  For example, when you are building a path a/b/c/d, the resolving shortcut 
code will never be hit  if a, b, and c exists.  

As for unix, symlink resolution is done by the kernl. 

  

Comment 7

[Matthew Tuck [:CodeMachine]]

M18 is an old milestone, moving forward in time.

Comment 8

[Doug Turner (:dougt)]

removing milestone.

Comment 9

[Doug Turner (:dougt)]

waterson, is there anyone on your team that can investigate the impl of nsLocalFile?

Comment 10

[Gagan]

dougt: is this the right milestone for this work? pls. retarget as necessary.

Comment 11

[Doug Turner (:dougt)]

untargeting as I don;t have the cycles. 

Comment 12

[Brendan Eich [:brendan]]

helpwanted, right?

/be

Comment 13

[Doug Turner (:dougt)]

cathleen, I am told that you have people that could look at this?  if not assign 
it back to me.

Comment 14

[Cathleen]

I was wondering if you can help take a look...  :-)

So, currently, there are a few things that raised the concers about nsIFiles 
during the performance meeting.  We don't have any quantified data, but people 
have heard that switching from nsFileSpec to nsIFiles, there are some 
performance issues.  Also, bug 71447 which we try to open 34 RDF files that 
don't exist takes 4% of startup time, increase the awareness of nsIFiles 
perforance.

If you can give me some directions to get me started, I'd be glad to do some 
investigations into it.  (I also dont have a Mac, so I need some help from mac 
people too!)

Comment 15

[Cathleen]

reassign back to you for now...  :-)

Comment 16

[chris hofmann]

lets put this on kandrot's list

Comment 17

[Simon Fraser [no longer active]]

kandrot says he's swamped, maybe try the original author, sdagley  :)

Comment 18

[Jussi-Pekka Mantere]

Moving to 0.9.3.

Comment 19

[Blake Ross]

evaluation won't finish tonight...

Comment 20

[Pete Collins (MDG)]

I can take performance evaluation for unix. 
I have already started writing a test suite for nsIFIle.

Taking bug.

--pete

Comment 21

[Simon Paquet [:sipaq]]

This bug is targeted at a Mac classic platform/OS, which is no longer supported
by mozilla.org. Please re-target it to another platform/OS if this bug applies
there as well or resolve this bug.

I will resolve this bug as WONTFIX in four weeks if no action has been taken.
To filter this and similar messages out, please filter for "mac_cla_reorg".

Comment 22

[Simon Paquet [:sipaq]]

This bug hasen't been touched for years and is clearly unowned. Moving back to default assignee/QA so that people, who are watching those can accurately triage this bug.

Also resetting Priority, Target Milestone, Status Whiteboard and Status.