17948 - UNIX nsIFile not implemented

Status: RESOLVED FIXED

Description

[leger]

Putting on PDT- radar.  Porkjockey work.  Not known why this is needed for
dogfood.

Comment 1

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

Rocking hard on this.  I'll mail out diffs and the Unix files for review today.

Comment 2

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

*** Bug 14360 has been marked as a duplicate of this bug. ***

Comment 3

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

Initial portions are in the tree.

Comment 4

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

shaver deftly dodges the oncoming M12 train.  He will catch his breath and leap
aboard the M13 one as soon as it starts boarding.

Comment 5

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

We've just got Move/Copy and isContainedIn that matter right now.

Comment 6

[Christopher Blizzard (:blizzard)]

Taking ownership from the sickly shaver.

Comment 7

[Mike Shaver (:shaver -- probably not reading bugmail closely)]

You rule, blizzard.

I've got isContainedIn implemented, but I'm crashing in autoreg now
(dirIterator->GetNext in nsNativeComponentLoader::RegisterComponentsInDir), so I
can't test it.  Foiled again!

I've got another build going now, in case that fixes it -- is anyone else seeing
this on the NSIFILE branch?

Comment 8

[chris hofmann]

m14

Comment 9

[Christopher Blizzard (:blizzard)]

Taking ownership and marking as beta1.  This is something that I would consider
important for beta quality.

Comment 10

[Warren Harris]

What's the status of this bug? As far as I know, we're running with nsIFile on 
linux (all of necko uses it now).

Comment 11

[Christopher Blizzard (:blizzard)]

There are still some things that aren't implemented yet.  The following
functions aren't done:

nsLocalFile::Normalize()
nsLocalFile::CopyTo() ( just copying a directory.  files work. )
nsLocalFile::CopyToFollowingLinks()
nsLocalFile::MoveTo()
nsLocalFile::GetDiskSpaceAvailable()
nsLocalFile::GetParent()
nsLocalFile::Spawn()

Comment 12

[leger]

Putting dogfood in the keyword field.

Comment 13

[Christopher Blizzard (:blizzard)]

nsIFile::Spawn() and nsIFile::Normalize have been checked in.

Comment 14

[Steve Morrison]

I have a version of GetParent as a patch over in bug 21100 - not sure if it's 
the best way, but, it's good enough for the filecache to start working.

Comment 15

[Christopher Blizzard (:blizzard)]

I'll take a look at it.  A brief glance says it's ok but I'll have to test it.

Comment 16

[Christopher Blizzard (:blizzard)]

Moving out...

Comment 17

[Christopher Blizzard (:blizzard)]

Please ignore the spam.  Changing address.

Comment 18

[Christopher Blizzard (:blizzard)]

bustage from my reassign

Comment 19

[Thord Andersson]

What is the status of this bug?
According to the comments it may be important for beta (nsbeta2 ?).

Comment 20

[leger]

Putting on [dogfood-] radar.  

Comment 21

[Christopher Blizzard (:blizzard)]

Moving out...

Comment 22

[Jason Eager]

Ok, doing a status update on this bug since the information given seems to be
WELL out of date.

Here is the current status:

::CopyTo is not implimented for directories.
::CopyToFollowingLinks is completely unimplimented.
::MoveTo is completely unimplimented.
::GetDiskSpaceAvailable is unimplimented for platforms without statfs.

Comment 23

[Daniel Veditz [:dveditz]]

The lack of MoveTo is blocking bug 37842 which we need to do XPI upgrades of 
the product itself, skins, etc.

Comment 24

[Christopher Blizzard (:blizzard)]

Yeah, I haven't had time to work on that at all.  Sorry. :(  Feel free to assign
it to dougt if you want.

Comment 25

[Doug Turner (:dougt)]

MoveTo() is implemented on all platforms now.  (Unix does a Copy/Delete - not
optimal)

Comment 26

[Brendan Eich [:brendan]]

Argh -- why does Unix not use rename(2) and only copy/unlink if rename fails
with EXDEV?

/be

Comment 27

[Christopher Blizzard (:blizzard)]

dougt admits that's it's not optimal probably for this very reason

Comment 28

[Brendan Eich [:brendan]]

Who's gonna fix it?  I can cough up a patch if no one else is gonna.

/be

Comment 29

[Daniel Veditz [:dveditz]]

Given that this bug has hung around unfixed so long, maybe you'd better just do 
it.

Comment 30

[Doug Turner (:dougt)]

remember that rename fails across volumes.  

Comment 31

[Christopher Blizzard (:blizzard)]

and that with recursive moves across volumes sometimes it will fail in the
middle if you happen to cross a filesystem border

Comment 32

[Brendan Eich [:brendan]]

Yeah yeah -- I mentioned EXDEV, didn't I?  (Meesa implemented rename in Irix 
some time in the late '80s.)

Blizzard, do you mind if I take this bug?

/be

Comment 33

[Christopher Blizzard (:blizzard)]

Sure, go right ahead.

Comment 34

[Brendan Eich [:brendan]]

Maybe I should take bug 33098 and close this one?

/be

Comment 35

[Doug Turner (:dougt)]

This has been fixed.  All apis are implemented.