Solaris10 TEST3 Build

[dijichi2]

Hi Jooray

Yes, I'll post detailed howto. The port I completed before has sort of stagnated because
a) svn tree stopped working so I couldn't track against my changes
b) I didn't produce patches and submit against bug reports so they couldn't be integrated
c) I stopped documented the fixes halfway along

I'll fix b) and c) as I go along this time, the documentation is important as is patches, so I'll post both and you can build from them.

As a change to earlier post, I'm waiting for 5.0b2 source to be released then I'll do a port against that.

Cheers
Dom

[epoc1000]

Hi,

I'm very interested in a Solaris 10 version of Zimbra, because I'm running a lot of servers with Solaris 10. So, I just wanted to know, if there is a roadmap for a stable build or a Zimbra supported version?

Bye
Sebastian

[qu1j0t3]

Yes, I'd echo Sebastian's question. I want to get a stable release (such as 4.5.5) running on Solaris 10 ASAP.

[dijichi2]

i can't answer for zimbra, but i will start providing more test versions from 5.0b2 onwards (no support for 4.x branch).

if you want supported versions of zimbra, you have to talk to their sales teams. the ports are largely driven by economics - the more people request to buy zimbra on a certain platform, the more incentive zimbra has to produce the port.

[epoc1000]

That's great to hear. I had no time to something my self for testing a Solaris 10 version, so maybe it would be helpfull to let us in your building process. So, everybody maybe can help u. I think the power of Solaris 10 (zones, zfs and so on...) and Zimbra is a great combination.

[qu1j0t3]

I'm still unclear. I get the impression that you succeeded with a version of 4.x? That recipe would be useful to me. I don't do betas.

Does anyone know what's happening with the Svn code drop? Why are there no tags in it, btw?

[dijichi2]

That's great to hear. I had no time to something my self for testing a Solaris 10 version, so maybe it would be helpfull to let us in your building process. So, everybody maybe can help u. I think the power of Solaris 10 (zones, zfs and so on...) and Zimbra is a great combination.

absolutely. the previous solaris builds i released, i didn't get feedback from anybody despite all the requests for solaris builds, so clearly i was going about it the wrong way. this next build I will release fully patched source tarballs and detailed instructions on how to build, maybe that will be more useful to people.

I'm still unclear. I get the impression that you succeeded with a version of 4.x? That recipe would be useful to me. I don't do betas.

Don't do zimbra on solaris then, any form of it is not even beta, best described as early-alpha. and not official at that, just a community hack. i will post the recipe for those interested, however. i did get it working on 4.x but as the svn tree is broken i have to start from scratch and i might as well do that with the 5.x tree.

Does anyone know what's happening with the Svn code drop? Why are there no tags in it, btw?

according to comments from zimbra on this forum, it's broken, has been broken for quite some months and they're working hard to get it fixed, hopefully sometime in the near future.

[qu1j0t3]

It's not a question of "not doing Zimbra on Solaris". I've spent a lot of time porting it already, a few months ago. I got all of third party ported and had to put it aside at the point of looking at the application itself (undoubtedly a smaller challenge). I have no objections to running a stable release on an unsupported O/S. But I see no point in getting involved in application issues (beta) in addition to platform issues.

At the time I did that work, the build system and documentation was half cooked. I hope it's improved.

(Porting third party essentially duplicates the effort of packagers - it's extra effort on all platforms, not just Solaris... IMHO it would be better if Z relied on standard packages for as much as possible.)

That said, I think running Zimbra is probably the best solution for my company, and if we run it on any O/S, it will be Solaris 10.

There is no way forward at the moment, even if I wanted to start over, without sensible access to Z source.

[epoc1000]

absolutely. the previous solaris builds i released, i didn't get feedback from anybody despite all the requests for solaris builds, so clearly i was going about it the wrong way. this next build I will release fully patched source tarballs and detailed instructions on how to build, maybe that will be more useful to people.

Maybe Zimbra itself will help to get it running on Solaris. If you need any help (svn, test system, docs or something else) please tell me. We are running most of our infrastructure on Solaris. The Zimbra servers are almost the only ones running on Linux.

[dijichi2]

It's not a question of "not doing Zimbra on Solaris". I've spent a lot of time porting it already, a few months ago. I got all of third party ported and had to put it aside at the point of looking at the application itself (undoubtedly a smaller challenge). I have no objections to running a stable release on an unsupported O/S. But I see no point in getting involved in application issues (beta) in addition to platform issues.

Good luck then. Vast majority of thirdparty is already written to run on solaris, so it's not really porting, more of a build process that needs defining - there is the odd exception to this eg. perdition. Contrary to what one would think from essentially a java app, Zimbra was paradoxically a much bigger challenge. I'll try and produce patches as well, perhaps you can apply them to 4.x tree.

I hear what you're saying about not adding app beta issues to the platform issues, but I took the approach that if my hackjob ever ends up contributing to an official port, it certainly wouldn't happen in 4.x series but might end up in 5.x. Often in opensource world betas are more stable than the last releases, and as Zimbra themselves use 5.x internally I took it that it was stable enough to make it worthwhile. As we can't track against svn I don't have time to do more than one port. Running Zimbra on a new platform takes iterations of trial and error before it would become stable enough for production use. I don't see this happening in the 4.x series.

At the time I did that work, the build system and documentation was half cooked. I hope it's improved.

It hasn't.

There is no way forward at the moment, even if I wanted to start over, without sensible access to Z source.

Well, does make it more difficult!