binary updates for firefox and chromium?

[Jan Beich]

Peter Vereshagin <peter at vereshagin.org> writes:

> Hello.
>
> 2013/01/31 18:44:30 +0000 Walter Hurry <walterhurry at gmail.com> => To
> freebsd-questions at freebsd.org :
>
> WH> What, for example, is the problem with compiling Firefox 18 from the port?
> WH> 
> WH> It takes about an hour to compile on my main box (9.1 Release on- 
> WH> x86_64), which has nothing special at all by way of hardware.
>
> It use to take me a day (~20h) to rebuild each of: www/firefox,
> www.seamonkey, mail/thunderbird.

We can certainly shave some time and maintenance burden by using a
shared libxul.so similar to how PkgSrc and Debian did. But the work
required is bit too big for me who has little motivation to hack outside
of firefox-nightly or upstream.

>
> In general I'm satisfied with speed of their work though.
>
> Starting from v17 upgrades I experience the problem I can't see the solution
> so I even don't ask about it. The details is that for all the 3 ports almost
> everything builds and then while linking the main (or almost the main) binary
> the compiler gets to lose the 'JSAutoCompartment.o'.

I'd recommend to build ports inside a poudriere jail or tinderbox.
Sorting out issues in the environment could be a huge timesink.
For one, I gave up doing so with VirtualBox or Chromium ports.

>
> I know I can buy some cloud to rebuild but is it of any rational if every
> hardware-like-mine owner shall follow this way when we can have it
> centralized?

Try bugging Redports people if they plan to provide user package
repositories similar to Launchpad. While one can share repos on maillist
they lack necessary metadata in order to trust such packages:

- an exact revision/source the package was built against
- build environment
- log/output of the build process
- a sort of proof that the above wasn't forged (popularity?)
- checksum or package signing
- encrypted connection

TBH, I don't have much hope for Pointyhat. It was damn slow even for
Tier1 archs before the security incident.