Good Evening, It has been a few years since I've configured a Windows system for building OpenAFS, but I'm taking a shot at it on the same system I'm hosting the Debian buildslaves on. I have no guesses as to what the typical build completion time would be like on my system (provided I can successfully re-create the compilation environment needed). I'll let y'all know how it goes.
Assuming I can get everything working and the build time is reasonable, it may be possible to get it added into the Gerrit-triggered builds. However, my system is hosted from my house over a consumer-grade internet connection. I cannot guarantee 99.99% uptime for this system. What is the maximum acceptable outage window for Gerrit-triggered builds? Regards, Nathan On Thu, Dec 24, 2015 at 2:44 PM, Jason Edgecombe <[email protected]> wrote: > Hi Kostas, > > Ideally, the new windows build slave needs to be online 24/7. If the > proposed windows build slave is only online for 12 hours/day, then we could > still have scheduled builds, but gerrit-triggered builds would likely be > unworkable because gerrit commits might have to wait for 12+ hours to > receive build status feedback. > > I have no experience with building OpenAFS on windows, so I can't speak to > the software requirements in the build environment. I know that this sounds > odd because I'm the 'buildbot owner', but each buildslave admin sets up the > build environment, so that knowledge lies with each individual buildslave > maintainer. I can provide advice on how to install/configure the buildbot > software and how to compile on Linux platforms. > > Here is a quick explanation of our best build slave options (in my > opinion) and the trade-offs for each: > * gerrit-triggered, always on (24x7) - ideal case, builds trigger on > every gerrit change, build status is pushed into the gerrit system and > marks a commit as 'Verified'. The gerrit status isn't published until all > slaves have reported back. builds are queued up if slaves are offline. > * scheduled - builds are triggered based on time, the build status only > appears on the buildbot status page. > > gerrit-triggered builders have an added requirement of a maximum on the > build time, but that maximum time is determined by the community of > developers. How long do you want to wait for feedback? > > builbot can also support a 'latent' build slave where the buildbot > software will start and stop a VM (libvirt) or a cloud VM (AWS) on-demand. > I have no previous experience with this type of build slave and that would > require some extra research and testing to pull off, along with the added > complexity. > > The buildbot software is very flexible. Builds can be initiated by a > variety of triggers (gerrit, scheduled, manual, web hook, git push). There > are a variety of options for build status updates: email, IRC, gerrit, etc. > > You can read more about our version of the buildbot software at > http://docs.buildbot.net/0.8.10/index.html > > Sincerely, > Jason > > > On 12/24/2015 04:02 AM, Kostas Liakakis wrote: > >> Ben, Jason >> >> Do you have an estimation of CPU resources required for a Windows >> buildslave? Do they have to be available 24/7 or would a say, 12h window >> suffice? Would a Pentium G3260 cut it or is an i-class CPU required to >> complete the build jobs within a useful timeframe? >> >> How hard is the VS2005 or VS2008 requirement? In the build environment >> instructions link Ben posted, VS2012 is mentioned as an option for >> AFSVER_CL. VS2012 is the oldest I can easily get my hands on through our >> Dreamspark subscription. I think if VS version can be tackled, the rest >> of the build requirements are of a non-issue. >> >> How does one BUILD a buildslave for Windows after setting up the build >> environment? Is there any rundown for this? >> >> Regards, >> >> >> -K. >> >> >> On 12/23/2015 11:47 PM, Benjamin Kaduk wrote: >> >>> Hi all, >>> >>> The automated buildbot builder for Windows that performs a test build of >>> each code submission to gerrit before that code can be merged into the >>> tree has been disabled; please see the forwarded message below. That >>> builder was previously donated by Jeffrey Altman and provided tremendous >>> value to the OpenAFS project over the years; thank you Jeffrey for having >>> donated these resources for so long. >>> >>> It is my opinion that in the absence of a gerrit-triggered windows >>> builder, the OpenAFS code for windows will very quickly bitrot and become >>> unusable. The development environment for Windows imposes constraints >>> not >>> found on Unix systems, and most of our developers are unfamiliar with >>> those constraints, so without a way to automatically enforce them before >>> merging commits, the windows development constraints will quickly be >>> violated. >>> >>> If you are interested in OpenAFS continuing to function on the Windows >>> platforms, I urge you to contribute resources towards a functioning >>> Windows buildbot builder. >>> >>> http://git.openafs.org/?p=openafs.git;a=blob;f=README-WINDOWS;h=4f9af252f51677dd49e32c4af4411bd96eed53c9;hb=HEAD >>> has some (probably somewhat outdated) instructions for setting up a >>> Windows build environment. Do note that the specific (old) versions of >>> various softwares are important: newer versions will not work in most >>> cases. I may be able to provide some assistance in setting up a build >>> environment; please post questions encountered while setting up such a >>> development environment to [email protected]. >>> >>> Thank you, >>> >>> Ben Kaduk >>> >>> >>> >>> ---------- Forwarded message ---------- >>> Date: Wed, 23 Dec 2015 16:17:20 -0500 >>> From: Jason Edgecombe <[email protected]> >>> To: [email protected] >>> Subject: [OpenAFS-devel] Window build slaves are disabled and removed >>> >>> Hi everyone, >>> >>> The only windows build slave has been disabled and removed from the >>> OpenAFS >>> buildbot system. If anyone wants to stand up a replacement, please >>> contact me. >>> >>> You may see the current buildbot status and variety of build slaves at >>> http://buildbot.openafs.org/waterfall >>> >>> Sincerely, >>> Jason >>> _______________________________________________ >>> OpenAFS-devel mailing list >>> [email protected] >>> https://lists.openafs.org/mailman/listinfo/openafs-devel >>> _______________________________________________ >>> OpenAFS-devel mailing list >>> [email protected] >>> https://lists.openafs.org/mailman/listinfo/openafs-devel >>> >>> > _______________________________________________ > OpenAFS-devel mailing list > [email protected] > https://lists.openafs.org/mailman/listinfo/openafs-devel >
