590112 - Fix Android cross-compiles on FRR

Status: RESOLVED FIXED

(Tamarin Graveyard :: Build Config, defect, P3)

Description

[Felix S. Klock II [:pnkfelix, :fklock]]

Steven recently updated his flashruntime-redux (FRR) branch in P4 so that it builds the usual targets.  But there remain incompatibilities between the tip of tamarin-redux (TR) and the Android portion of FRR.  This ticket is intended to track the fixes necessary to TR in order to build and run an Android player.

(There are also patches necessary for FRR itself, but those are not meant to be directly posted here.)

Comment 1

[Felix S. Klock II [:pnkfelix, :fklock]]

(Sorry, I failed to follow proper process on P4 CL 707419 which changes FRR.  Its an obvious fix to a real problem.  But as a fan of proper process, I am ashamed; won't happen again.)

On a semi-related note, we may want to follow up and find out whether the Player's Android build infrastructure should be sym-linked into the makefiles in tamarin-redux, rather than trying to keep these files in sync...

Comment 2

[Felix S. Klock II [:pnkfelix, :fklock]]

(In reply to comment #1)
> (Sorry, I failed to follow proper process on P4 CL 707419 which changes FRR. 
> Its an obvious fix to a real problem.  But as a fan of proper process, I am
> ashamed; won't happen again.)

(by proper process, I mean "getting another human to review the patch")

Comment 3

[Felix S. Klock II [:pnkfelix, :fklock]]

Attachment: hack API with method to clear gc of a GCAutoEnter

This patch was the simplest way to get Android building against FRR.  It is probably not the right thing for the long-term; that is, it probably should not be pushed to the repository.  But I figured it was easiest to upload the patch here so that others who want to target Android can get something going quickly.

Comment 4

[Felix S. Klock II [:pnkfelix, :fklock]]

Attachment: kills dead (debug-only) private method

works around gcc warning turned error in the Android build.

When I discussed this with Lars some time ago [1], he mentioned the option of disabling the particular warning rather than changing the code to accommodate gcc.

But it really is a pretty simple mod, so maybe we should push it anyway...

[1] http://asteam.corp.adobe.com/irc/log/ttlogger/tamarin/20100818

Comment 5

[Steven Johnson]

Comment on attachment 469027 [details] [diff] [review]
kills dead (debug-only) private method

obsoleted by fix for https://bugzilla.mozilla.org/show_bug.cgi?id=598424

Comment 6

[Felix S. Klock II [:pnkfelix, :fklock]]

The remaining work should be covered by bug 598422.

Comment 7

[Felix S. Klock II [:pnkfelix, :fklock]]

Bug 598422 was resolved as "wontfix", which does not fix this bug, so I am reopening this bug.

(Tommy is in contact with the Android team about the issue; I'm just keeping notes/patches with this bug and want to keep its state accurate.)

Comment 8

[Trevor Baker]

is this still in play?

Comment 9

[Felix S. Klock II [:pnkfelix, :fklock]]

(In reply to comment #8)
> is this still in play?

yes it is, but I do not know how soon it will be resolved.  As recently as Friday (December 3rd), Tommy and Simon and I were chatting about it on the IRC channel.

See: http://asteam.corp.adobe.com/irc/log/ttlogger/tamarin/20101203
and scroll down to 15:55

Comment 10

[Felix S. Klock II [:pnkfelix, :fklock]]

Okay at some point the SetGCToNull went away.  :)