383833 - eliminate toolkit's mork-based history implementation

Status: RESOLVED FIXED

(Firefox :: Bookmarks & History, defect)

Description

[Robert Kaiser]

I just realized that toolkit's mork-based history implementation is currently unused, and we should not create new users for it, so best is to just remove it.

This allows us to not only remove a few files from cvs (toolkit/components/history/src/*) but also use MOZ_PLACES as a clean switch for migrating SeaMonkey and Camino from xpfe history to places history.

Comment 1

[Robert Kaiser]

Attachment: build history/public for places, never build all of history

This removes everything but history/public from build, along with that, the following files can be cvs removed:

mozilla/toolkit/components/history/.cvsignore
mozilla/toolkit/components/history/Makefile.in
mozilla/toolkit/components/history/src/.cvsignore
mozilla/toolkit/components/history/src/Makefile.in
mozilla/toolkit/components/history/src/nsGlobalHistory.cpp
mozilla/toolkit/components/history/src/nsGlobalHistory.h
mozilla/toolkit/components/history/src/nsHistoryLoadListener.h
mozilla/toolkit/components/history/src/nsMdbPtr.h

Comment 2

[Asaf Romano (gone)]

Comment on attachment 267794 [details] [diff] [review]
build history/public for places, never build all of history

I think the correct fix is to cvs copy the interface and then cvs remove toolkit/components/history  altogether. That said, non-places Fx builds are still supported... it might be just the right time to change that though.

Comment 3

[Robert Kaiser]

Is it OK to go for cvs copying the interface right now, make it be built from places accordingly, and apply the following hunk to the components/Makefile.in?

 ifndef MOZ_SUITE
 # XXX Suite doesn't want these just yet
 ifdef MOZ_XUL
 DIRS += \
        autocomplete \
-       history \
        passwordmgr \
        satchel \
        $(NULL)
 endif # MOZ_XUL
+ifndef MOZ_PLACES
+DIRS +=        history
+endif # MOZ_PLACES
 endif # MOZ_SUITE

Comment 4

[Robert Kaiser]

Attachment: step 1: only build history/ for non-places builds

This patch goes the first step by isolating history/ so that it is only built for non-places builds, as the interface places needed from there is now in places/public.

post-alpha6 we can go and remove this ifdef along with the history/ subdir, but until then, I was informed that we want to still support non-places builds.

Comment 5

[Robert Kaiser]

Attachment: step 1, v1.1: build history/ only for non-places builds, make places use its own nsIBrowserHistory.idl

Umm, sorry, of course we also need to do something with the copied IDL file ;-)

Comment 6

[Robert Kaiser]

Attachment: step 1, v1.2: build history/ only for non-places builds, use nsIBrowserHistory.idl from places (checked in)

...and sorry again, I found an additional change we need. This has been _really_ tested now, though, and it really compiles correctly.

Comment 7

[Asaf Romano (gone)]

Comment on attachment 269512 [details] [diff] [review]
step 1, v1.2: build history/ only for non-places builds, use nsIBrowserHistory.idl from places (checked in)

r=mano

Comment 8

[Robert Kaiser]

Comment on attachment 269512 [details] [diff] [review]
step 1, v1.2: build history/ only for non-places builds, use nsIBrowserHistory.idl from places (checked in)

This patch is checked in now, leaving the bug open for later completely removing the old history impl.

Comment 9

[(not reading, please use [e-mail] instead)]

before we remove the old implementation, I think we'd like to do some "old" vs "new" performance testing, right dietrich?

Comment 10

[Dietrich Ayala (:dietrich)]

(In reply to comment #9)
> before we remove the old implementation, I think we'd like to do some "old" vs
> "new" performance testing, right dietrich?
> 

yes, we're doing some places vs non-places testing under alice's new test harness in bug 374532.

Comment 11

[Igor Velkov]

why did bug 386392 and bug 374532 is blockers for this?

Comment 12

[Joshua Cranmer [:jcranmer]]

Seeing as how the consensus for bug 374532 is that a) there is a 2% regression for empty profiles and b) this regression is acceptable, would we be able to move forward on removing this now?

Comment 13

[Robert Kaiser]

Attachment: step 2: finally kill toolkit history

This is the final second step to actually kill off toolkit's mork-based history.

Comment 14

[Asaf Romano (gone)]

Comment on attachment 335390 [details] [diff] [review]
step 2: finally kill toolkit history

r=mano

Comment 15

[Robert Kaiser]

Pushed as http://hg.mozilla.org/mozilla-central/rev/428c66910e66

Comment 16

[Florian Quèze [:florian]]

I suspect you may also want to remove the line "tkhstory" (currently line 266) and maybe all the else branch after ifdef MOZ_PLACES in toolkit/library/libxul-config.mk