[opensuse-buildservice] Unneeded build loop
Hi! Yesterday I was experiencing some problems with the build service. I don't know if it was really slow, or it was a connection problem from my part. The fact is that I submitted the spec files many times, and triggered a rebuild on each upload, which, let's say, created 5 build requests. Today I woke up and found out that the build service is still trying to build my packages, after my erroneous 5 triggers. The question is: I think there is no need to schedule a rebuild if you already scheduled one before. IMHO, the second rebuild request should either (a) override the first one or (b) be dropped until the first one is finished. I vote for (a), since people make mistakes and it's very probable that the second request is the right one. I guess that way we would spare some building time on the service. WDYT? -- % Mauricio Teixeira (netmask) % mteixeira{a}webset{d}net <> Maceio/AL/BR % http://mteixeira.webset.net <> http://pmping.sf.net
On Thu 10 Aug 2006 13:00, Mauricio Teixeira (netmask) wrote:
Hi!
Yesterday I was experiencing some problems with the build service. I don't know if it was really slow, or it was a connection problem from my part.
The fact is that I submitted the spec files many times, and triggered a rebuild on each upload, which, let's say, created 5 build requests. Today I woke up and found out that the build service is still trying to build my packages, after my erroneous 5 triggers.
The question is: I think there is no need to schedule a rebuild if you already scheduled one before. IMHO, the second rebuild request should either (a) override the first one or (b) be dropped until the first one is finished. I vote for (a), since people make mistakes and it's very probable that the second request is the right one.
I guess that way we would spare some building time on the service.
I have had this problem before also. Definitely option A is the correct one. -- Peter Nixon http://www.peternixon.net/ PGP Key: http://www.peternixon.net/public.asc
On 8/10/06, Peter Nixon <listuser@peternixon.net> wrote:
either (a) override the first one or (b) be dropped until the first one is finished. I vote for (a), since people make mistakes and it's very probable that the second request is the right one.
I guess that way we would spare some building time on the service.
I have had this problem before also. Definitely option A is the correct one.
I think it may be queuing for build for every uploaded file too, so if we upload two changed files (spec and source tarball) it would build twice. I may be wrong though. It surely behaves that way uploading via web interface. Thanks to buildservice irc (darix?) I no longer upload files that way. It would also be useful for newbies like me to cancel the builds half way if we realize we've made a mistake or is scheduled for a very long time. Jigish PS. Sorry, the reply didn't go to the list the first time.
On Thu, Aug 10, 2006 at 03:48:02PM +0530, Jigish Gohil wrote:
I think it may be queuing for build for every uploaded file too, so if we upload two changed files (spec and source tarball) it would build twice. I may be wrong though.
No, that's right. We need to change the web interface to allow multiple uploads and a commit afterwards. Work in progress...
It surely behaves that way uploading via web interface. Thanks to buildservice irc (darix?) I no longer upload files that way.
It would also be useful for newbies like me to cancel the builds half way if we realize we've made a mistake or is scheduled for a very long time.
It's not easy to cancel running builds (scheduled ones are no problem, though). I'll try to come up with something. Cheers, Michael. -- Michael Schroeder mls@suse.de main(_){while(_=~getchar())putchar(~_-1/(~(_|32)/13*2-11)*13);}
On Thu 10 Aug 2006 13:26, Michael Schroeder wrote:
On Thu, Aug 10, 2006 at 03:48:02PM +0530, Jigish Gohil wrote:
I think it may be queuing for build for every uploaded file too, so if we upload two changed files (spec and source tarball) it would build twice. I may be wrong though.
No, that's right. We need to change the web interface to allow multiple uploads and a commit afterwards. Work in progress...
Yep. I always try to commit source, patches and spec in one go with osc but I invariable forget something or other and have to do a second commit, then wait for the first build to fail :-(
It surely behaves that way uploading via web interface. Thanks to buildservice irc (darix?) I no longer upload files that way.
It would also be useful for newbies like me to cancel the builds half way if we realize we've made a mistake or is scheduled for a very long time.
It's not easy to cancel running builds (scheduled ones are no problem, though). I'll try to come up with something. Cool
-- Peter Nixon http://www.peternixon.net/ PGP Key: http://www.peternixon.net/public.asc
Am Thursday 10 August 2006 12:38 schrieb Peter Nixon:
On Thu 10 Aug 2006 13:26, Michael Schroeder wrote:
On Thu, Aug 10, 2006 at 03:48:02PM +0530, Jigish Gohil wrote:
I think it may be queuing for build for every uploaded file too, so if we upload two changed files (spec and source tarball) it would build twice. I may be wrong though.
No, that's right. We need to change the web interface to allow multiple uploads and a commit afterwards. Work in progress...
Yep. I always try to commit source, patches and spec in one go with osc but I invariable forget something or other and have to do a second commit, then wait for the first build to fail :-(
you can also test your build before via "osc build" on your local system. So there is no reason why you need to wait for the server. (exception is that BuildRequires: changs needs to commited atm, will be fixed after the API change). bye adrian -- Adrian Schroeter SUSE Linux Products GmbH, Maxfeldstr. 5, 90409 Nuernberg, Germany email: adrian@suse.de
Em Qui, 2006-08-10 às 15:07 +0200, Adrian Schröter escreveu:
you can also test your build before via "osc build" on your local system. So there is no reason why you need to wait for the server.
Excuse me if I sound rude, but, I, personally, can't see the need of the build server if I had to build locally first. :) No harm intended. -- % Mauricio Teixeira (netmask) % mteixeira{a}webset{d}net <> Maceio/AL/BR % http://mteixeira.webset.net <> http://pmping.sf.net
Hello, "Mauricio Teixeira (netmask)" <netmask@webset.net> [2006-08-10]:
Em Qui, 2006-08-10 às 15:07 +0200, Adrian Schröter escreveu:
you can also test your build before via "osc build" on your local system. So there is no reason why you need to wait for the server.
Excuse me if I sound rude, but, I, personally, can't see the need of the build server if I had to build locally first. :)
Maybe you have no x86_64 machine at home? Maybe you don't want to rebuild factory every time some library changes? Maybe you don't want to rebuild kernel modules every time an update kernel is released? There are several reasons. :) Regards, Bernhard -- Die Arbeit ist ein Segen, der wie ein Fluch aussieht. -- Paul Auster
On 2006-08-10 16:59:39 -0300, Mauricio Teixeira (netmask) wrote:
Em Qui, 2006-08-10 às 15:07 +0200, Adrian Schröter escreveu:
you can also test your build before via "osc build" on your local system. So there is no reason why you need to wait for the server.
Excuse me if I sound rude, but, I, personally, can't see the need of the build server if I had to build locally first. :)
No harm intended.
netmask... of course you can do the "upload -> wait -> fail" loop on the build server. but building at least _once_ locally sounds smart to me. so you at least know that the package, you are about to submit, passed at least on one distro/arch combination. noone asked you to build for all targets before. if that is too much for you ... and you still dont see the benefit of the buildservice ... i dont know what to say. darix -- openSUSE - SUSE Linux is my linux openSUSE is good for you www.opensuse.org
On Thu, Aug 10, 2006 at 10:44:51PM +0200, Marcus Rueckert wrote:
On 2006-08-10 16:59:39 -0300, Mauricio Teixeira (netmask) wrote:
Em Qui, 2006-08-10 às 15:07 +0200, Adrian Schröter escreveu:
you can also test your build before via "osc build" on your local system. So there is no reason why you need to wait for the server.
Excuse me if I sound rude, but, I, personally, can't see the need of the build server if I had to build locally first. :)
No harm intended.
netmask... of course you can do the "upload -> wait -> fail" loop on the build server. but building at least _once_ locally sounds smart to me. so you at least know that the package, you are about to submit, passed at least on one distro/arch combination. noone asked you to build for all targets before.
if that is too much for you ... and you still dont see the benefit of the buildservice ... i dont know what to say.
Please, folks, we don't care about how the users of the build service work. I don't mind uploading without testing. This is a policy issue, i.e. multiple users working on one project might decide that the uploaded packages have to compile. But it is a local decision, not something we have to enforce globally. Cheers, Michael. -- Michael Schroeder mls@suse.de main(_){while(_=~getchar())putchar(~_-1/(~(_|32)/13*2-11)*13);}
Am Donnerstag, 10. August 2006 22:49 schrieb Michael Schroeder:
if that is too much for you ... and you still dont see the benefit of the buildservice ... i dont know what to say.
Please, folks, we don't care about how the users of the build service work. That sentence might be easily missunderstood ;-) but I agree fully in this context.
I don't mind uploading without testing. Right. We can't avoid so we have to accept it :-)
Klaas
Em Qui, 2006-08-10 às 22:44 +0200, Marcus Rueckert escreveu:
if that is too much for you ... and you still dont see the benefit of the buildservice ... i dont know what to say.
I use the build service so I don't need to have all the necessary packages to compile locally. It's just a matter of option, but sure your way sounds much more polite for people that build lots of packages. ;) -- % Mauricio Teixeira (netmask) % mteixeira{a}webset{d}net <> Maceio/AL/BR % http://mteixeira.webset.net <> http://pmping.sf.net
On 2006-08-10 22:06:13 -0300, Mauricio Teixeira (netmask) wrote:
Date: Thu, 10 Aug 2006 22:06:13 -0300 From: "Mauricio Teixeira (netmask)" <netmask@webset.net> Subject: Re: [opensuse-buildservice] Re: Unneeded build loop To: opensuse-buildservice@opensuse.org X-Mailer: Evolution 2.6.0
Em Qui, 2006-08-10 às 22:44 +0200, Marcus Rueckert escreveu:
if that is too much for you ... and you still dont see the benefit of the buildservice ... i dont know what to say.
I use the build service so I don't need to have all the necessary packages to compile locally.
It's just a matter of option, but sure your way sounds much more polite for people that build lots of packages.
"osc build repos arch yourspec" it will download the needed rpms into a local cache dir and starts "build". build (a script) sets up a chroot and runs the rpmbuild process inside. you can build with the same magic as the build service locally. darix -- openSUSE - SUSE Linux is my linux openSUSE is good for you www.opensuse.org
On 2006-08-10 15:48:02 +0530, Jigish Gohil wrote:
Date: Thu, 10 Aug 2006 15:48:02 +0530 From: Jigish Gohil <jigish.gohil@gmail.com> Subject: Re: [opensuse-buildservice] Re: Unneeded build loop Cc: opensuse-buildservice@opensuse.org
On 8/10/06, Peter Nixon <listuser@peternixon.net> wrote:
either (a) override the first one or (b) be dropped until the first one is finished. I vote for (a), since people make mistakes and it's very probable that the second request is the right one.
I guess that way we would spare some building time on the service.
I have had this problem before also. Definitely option A is the correct one.
I think it may be queuing for build for every uploaded file too, so if we upload two changed files (spec and source tarball) it would build twice. I may be wrong though.
It surely behaves that way uploading via web interface. Thanks to buildservice irc (darix?) I no longer upload files that way.
It would also be useful for newbies like me to cancel the builds half way if we realize we've made a mistake or is scheduled for a very long time.
use "osc" and not the webfrontend. darix -- openSUSE - SUSE Linux is my linux openSUSE is good for you www.opensuse.org
Am Donnerstag, 10. August 2006 12:37 schrieb Marcus Rueckert:
use "osc" and not the webfrontend.
I can not understand this statement. Do you think we should stop developing the webfrontend? My understanding is that all clients should work correctly and non should be more functional than another. So I go with Michael in another mail where he says: "that's a missing webclient feature that will come."
Klaas
participants (8)
-
Adrian Schröter -
Bernhard Walle -
Jigish Gohil -
Klaas Freitag -
Marcus Rueckert -
Mauricio Teixeira (netmask) -
Michael Schroeder -
Peter Nixon