[gccsdk] FW: Autobuilder packaging progress

alan buckley alan_baa at hotmail.com
Mon Jan 14 11:30:49 GMT 2008


> On Sun, 13 Jan 2008 21:00:36 -0800 Peter Naulls wrote:
>
> In message 
> alan buckley  wrote:
>
> [imagemagick]
>
>> You've beaten me to it as it didn't build on cygwin for me yesterday,
>> so I'm sure I was a close second!
>
> I'd like to hear of any failures of course. However, there were a
> number of transient issues forcing static builds, and dozens of fixes
> all over the place since you wrote this, so it might be just fine.

I've a few more little details to deal with in my current setup, but
I'll hopefully manage to move to GCC4 in the next week or so and
I'll try again then.

>
>> The current website is built from my cygwin machine. So the easiest
>> thing would be to send me the package and source zip files so I can
>> include them in the next website build.
>>
>> The build website can be run from anywhere as it just scans the
>> autobuilder_packages directory and creates a new directory with
>> the website. The only advantage of using my machine is that I
>> can diff it with my last build and only upload things that have
>> changed.
>
> I'm not sure of the precise mechanics (but I'll figure it out very
> shortly), but I expect it'll be something along the lines of a package
> being SCPed to a certain location (by you, John or myself as the main
> protangonists here, and the only ones with access) and having the index
> automatically regenerated on the hosting machine every hour or so.

I don't think the index needs to be regenerate so often. Thinking about
what you've said here, would it be possible for us to SCP the package and
its source to a staging area where a process could check for additions
and then schedule an early morning rebuild of the website and package
index. All the rebuild would need to do is copy the changes to the
autobuilder_packages directory on this machine and run the build-website
script.

>
> Incidentally, there's a bug with the 64-bit build of pkgindex. I didn't
> look too closely, but it gets stuck in an infinite loop writing the
> package index - somewhere in pkg::operator<< perhaps. Fortunately, I
> could compile and use the 32-bit version on my machine directly.
>
I think I can see what the problem may be here. Obviously I can't test
it but I'll check in a possible solution later today for you to try.

Regards,
Alan

_________________________________________________________________
Who's friends with who and co-starred in what?
http://www.searchgamesbox.com/celebrityseparation.shtml


More information about the gcc mailing list