1121217 - Add a configure flag when building the b2g runtime for browser.html

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Runtime, defect)

Comment 1

[[:fabrice] Fabrice Desré]

Attachment: browser-html-config.patch

Comment 2

[Mike Hommey [:glandium]]

Comment on attachment 8548484 [details] [diff] [review]
browser-html-config.patch

Review of attachment 8548484 [details] [diff] [review]:
-----------------------------------------------------------------

I know the context, but that doesn't tell me why it's needed. Especially, it's completely disconnected from anything else related. How is MOZ_BROWSER_HTML used? Where would --enable-browser-html be set?

Comment 3

[[:fabrice] Fabrice Desré]

browser.html will reuse the b2g runtime, but we need a different behavior in some cases:
- different defaults prefs.
- custom code under b2g/ that is only for browser.html (because the way we startup is different).
- disabling MOZ_UA_OS_AGNOSTIC

and probably others that will come up with time.

We will turn on the flag when we set up the releng parts to create builds.

Comment 4

[Mike Hommey [:glandium]]

To me, it seems like you should have a --enable-application=browser-html or whatever codename you want to use, and if you need to build b2g stuff from there, well, you can add DIRS += ['../b2g'].

Comment 5

[[:fabrice] Fabrice Desré]

Attachment: browserhtml-product.patch

After discussing on irc, we decided to create a new product, that builds like b2g for now but with its own confvars.sh and config/mozconfigs.

I can't push to try right now, will do as soon as try reopens.

Comment 6

[[:fabrice] Fabrice Desré]

Attachment: htmlrunner-product.patch

Same patch except that I renamed the product to "htmlrunner".

Comment 7

[Mike Hommey [:glandium]]

Comment on attachment 8550081 [details] [diff] [review]
htmlrunner-product.patch

Review of attachment 8550081 [details] [diff] [review]:
-----------------------------------------------------------------

::: b2g/confvars.sh
@@ +7,5 @@
>  
>  MOZ_APP_VERSION=38.0a1
>  MOZ_APP_UA_NAME=Firefox
>  
> +MOZ_B2G=1

It's already set on line 57.

::: htmlrunner/app.mozbuild
@@ +1,1 @@
> +# vim: set filetype=python:

htmlrunner sounds like webrunner and seems over-generic. browser.html had the benefit on making it clear that it was a browser.

::: htmlrunner/config/mozconfigs/common
@@ +12,5 @@
> +if test -d $topsrcdir/../gcc/bin; then
> +    HOST_CC="$topsrcdir/../gcc/bin/gcc"
> +    HOST_CXX="$topsrcdir/../gcc/bin/g++"
> +    ac_add_options --enable-stdcxx-compat
> +fi

This block is gonk specific. I doubt you'll need it.

::: htmlrunner/config/mozconfigs/linux32_gecko/debug
@@ +1,1 @@
> +. "$topsrcdir/b2g/config/mozconfigs/common"

seems to me this shouldn't be linux32_gecko, but someone from releng would tell for good. This also begs the question: do you really intend this to run on automation? Come to think of it, this does seem like a waste of resources, for something that is experimental. Wouldn't it be better for everyone if you could just download a b2g desktop build, edit a file, and run browser.html with it?

@@ +16,5 @@
> +
> +# Needed to enable breakpad in application.ini
> +export MOZILLA_OFFICIAL=1
> +
> +export MOZ_TELEMETRY_REPORTING=1

maybe not?

@@ +31,5 @@
> +ENABLE_MARIONETTE=1
> +export CXXFLAGS=-DMOZ_ENABLE_JS_DUMP
> +
> +# Include Firefox OS fonts.
> +MOZTTDIR=$topsrcdir/moztt

really?

::: htmlrunner/confvars.sh
@@ +7,5 @@
> +
> +MOZ_B2G=1
> +MOZ_HTMLRUNNER=1
> +
> +MOZ_APP_VERSION=38.0a1

Why not use $FIREFOX_VERSION? (b2g could do the same btw)

@@ +30,5 @@
> +NSS_DISABLE_DBM=1
> +MOZ_NO_EV_CERTS=1
> +MOZ_DISABLE_EXPORT_JS=1
> +
> +if test "$OS_TARGET" = "Android"; then

I doubt you'll ever build this.

@@ +39,5 @@
> +
> +# use custom widget for html:select
> +MOZ_USE_NATIVE_POPUP_WINDOWS=1
> +
> +if test "$LIBXUL_SDK"; then

no need to support LIBXUL_SDK, that option doesn't work anymore anyways.

@@ +47,5 @@
> +fi
> +
> +MOZ_MEDIA_NAVIGATOR=1
> +
> +MOZ_APP_ID={3c2e2abc-06d4-11e1-ac3b-374f68613e61}

seems to me you should use a different id.

Comment 8

[[:fabrice] Fabrice Desré]

Closing for now as I think we finally won't need that once we land bug 1118138.

Comment 9

[[:fabrice] Fabrice Desré]

Reopening as tweaking the current b2g will really be ugly. htmlrunner patch coming.

Comment 10

[[:fabrice] Fabrice Desré]

Attachment: htmlrunner-product.patch

Updated patch. I'm checking with releng to make builds happen.

Comment 11

[Paul Rouget [:paul]]

Not sure if it's relevant for this patch or not, but do we want to keep /b2g/chrome code in htmlrunner? Maybe we want htmlrunner to provide its own shell code.

Comment 12

[[:fabrice] Fabrice Desré]

Attachment: graphene.patch

Alright, after quite some discussions, we decided to go with a simple flag (--enable-graphene) and use that in b2g/ to tweak whatever we need.

I added the mozconfigs according to what catlee agreed on to do builds.

Comment 13

[[:fabrice] Fabrice Desré]

https://hg.mozilla.org/projects/larch/rev/2227b1f18a5d

Comment 14

[Paul Rouget [:paul]]

Fabrice, in b2g/app/macbuild/Contents/Info.plist.in, b2g should become graphene. Using %APP_BINARY% should do it.

Comment 15

[Paul Rouget [:paul]]

Attachment: fix plist

This is required to build a valid mac package.

Comment 16

[[:fabrice] Fabrice Desré]

Thanks, pushed at https://hg.mozilla.org/projects/larch/rev/d81cfcf097d3

Comment 17

[Kevin Grandon :kgrandon]

Landed in mozilla-central via bug 1204965.