Re: FindBugs exclusion policy proposal (was: Re: junit tests in nightly builds etc.)
I think the existing exclusions should be left in trunk, and that no new ones should be permitted in (or should be fixed immediately). If you do as you suggest below, then the list of findbugs errors will just continue to grow because nobody will pay attention to them. We are at a known, stable point, we do have some exclusions that we know need fixing, and we can do that as time permits; but let's keep it that way and not backpedal by allowing in new ones. G. On Thu, Feb 24, 2011 at 8:40 AM, Andreas Delmelle < andreas.delme...@telenet.be> wrote: > > No response to any of the posts in particular, just a general > thought/proposal. > > I can appreciate that the ComplexScripts branch requires a clean FB report > so that Glenn is not continuously sent on a wild goose chase. > However, personally (and Vincent seems to agree), I do not favor 'blind' > exclusions just to make the warnings go away. Following the same reasoning, > we could define thousands of CheckStyle suppressions, instead of encouraging > people to do it correctly. > > I do not have a problem with looking into those issues, if no one else has > the time and/or motivation, although that will not always happen > _immediately_. > > The general idea is good, but I am wondering, given the circumstances, if > we had not better invert the approach: keep the warnings alive in trunk, and > add exclusions for them only in the branch. > That way, devs who are not involved in the branch but do use FB, will be > constantly reminded that those issues should be looked into. For the > maintainer(s) of the branch, if the exclusion is properly commented, it can > serve as an indication that the warning originated in trunk and has nothing > to do with their changes. Should a genuine bug result from it, and it turns > out to hamper the development on the branch, it can then be raised as a > priority issue on this list. > > Ultimately, it is still a worthwhile goal to eliminate all of the warnings, > but we also have to be realistic enough to admit that that will not happen > overnight. > > > WDYT? > > Regards, > > Andreas > --- > >
FindBugs exclusion policy proposal (was: Re: junit tests in nightly builds etc.)
No response to any of the posts in particular, just a general thought/proposal. I can appreciate that the ComplexScripts branch requires a clean FB report so that Glenn is not continuously sent on a wild goose chase. However, personally (and Vincent seems to agree), I do not favor 'blind' exclusions just to make the warnings go away. Following the same reasoning, we could define thousands of CheckStyle suppressions, instead of encouraging people to do it correctly. I do not have a problem with looking into those issues, if no one else has the time and/or motivation, although that will not always happen _immediately_. The general idea is good, but I am wondering, given the circumstances, if we had not better invert the approach: keep the warnings alive in trunk, and add exclusions for them only in the branch. That way, devs who are not involved in the branch but do use FB, will be constantly reminded that those issues should be looked into. For the maintainer(s) of the branch, if the exclusion is properly commented, it can serve as an indication that the warning originated in trunk and has nothing to do with their changes. Should a genuine bug result from it, and it turns out to hamper the development on the branch, it can then be raised as a priority issue on this list. Ultimately, it is still a worthwhile goal to eliminate all of the warnings, but we also have to be realistic enough to admit that that will not happen overnight. WDYT? Regards, Andreas ---
Re: junit tests in nightly builds
On Wed, Feb 23, 2011 at 01:10:10PM -0700, Glenn Adams wrote: > Right now the nightly build is our CI process. That is a development-centric point of view. The nightly build is definitely not a CI process. It is a service to the users, which must not be interrupted by development concerns. If a reliable application can be built, it must be built and delivered. Simon
Re: junit tests in nightly builds
Well, not quite: http://vmgump.apache.org/gump/public/xml-fop/ Basics work but it needs some love apparently. On 23.02.2011 21:10:10 Glenn Adams wrote: > Right now the nightly build is our CI process. > > On Wednesday, February 23, 2011, Vincent Hennebert > wrote: > > On 23/02/11 14:42, Glenn Adams wrote: > >> I guess we disagree, since I believe application quality and code quality > >> are related. And, further, I believe findbugs at least can identify real, > >> functional bugs (as opposed to checkstyle). > > > > While I agree with the above, Im with Simon on this. Tests should be > > done within a continuous integration tool; Nightly builds serve > > a different purpose. > > > > I increasingly feel the need to set up continuous integration for the > > FOP project. The ASF provides several CI environments (Hudson, among > > others), at some point in the future Im going to try them out and set > > up something. Hopefully sooner rather than later. > > > > > > Vincent > > > > > >> G. > >> > >> On Wed, Feb 23, 2011 at 2:01 AM, Simon Pepping > >> wrote: > >> > >>> On Wed, Feb 23, 2011 at 01:15:34AM -0700, Glenn Adams wrote: > OK, understand on the junit headless issue. For checkstyle/findbugs it > >>> would > be useful to fail the nightly build if they do not pass. I will > >>> investigate > the necessary changes to enable this option, which I hope can be adopted. > >>> > >>> I would not agree. Nightly builds are a courtesy to the user. It would > >>> be good if we could guarantee that the builds pass the junit tests. > >>> But it is not relevant to the user whether they pass checkstyle and > >>> findbugs rules. These tests address the issue of code quality, not of > >>> application quality. > >>> > >>> Simon > >>> > On Wed, Feb 23, 2011 at 12:55 AM, Simon Pepping wrote: > > > On Tue, Feb 22, 2011 at 11:25:20AM -0700, Glenn Adams wrote: > >> I notice also that the nightly build target does not run all the > >>> junit > >> tests. It would be better if it run all of them plus checkstyle and > >> findbugs. > > > > Many junit tests require a display. Nightly builds are run in a > > headless configuration, hence I had to disable many junit tests. At > > nightly builds there is no one to check checkstyle and findbugs errors > > and warnings; therefore there is no point in running them. > > > > Simon > > > >>> > >> > > Jeremias Maerki
Re: junit tests in nightly builds
Right now the nightly build is our CI process. On Wednesday, February 23, 2011, Vincent Hennebert wrote: > On 23/02/11 14:42, Glenn Adams wrote: >> I guess we disagree, since I believe application quality and code quality >> are related. And, further, I believe findbugs at least can identify real, >> functional bugs (as opposed to checkstyle). > > While I agree with the above, I’m with Simon on this. Tests should be > done within a continuous integration tool; Nightly builds serve > a different purpose. > > I increasingly feel the need to set up continuous integration for the > FOP project. The ASF provides several CI environments (Hudson, among > others), at some point in the future I’m going to try them out and set > up something. Hopefully sooner rather than later. > > > Vincent > > >> G. >> >> On Wed, Feb 23, 2011 at 2:01 AM, Simon Pepping wrote: >> >>> On Wed, Feb 23, 2011 at 01:15:34AM -0700, Glenn Adams wrote: OK, understand on the junit headless issue. For checkstyle/findbugs it >>> would be useful to fail the nightly build if they do not pass. I will >>> investigate the necessary changes to enable this option, which I hope can be adopted. >>> >>> I would not agree. Nightly builds are a courtesy to the user. It would >>> be good if we could guarantee that the builds pass the junit tests. >>> But it is not relevant to the user whether they pass checkstyle and >>> findbugs rules. These tests address the issue of code quality, not of >>> application quality. >>> >>> Simon >>> On Wed, Feb 23, 2011 at 12:55 AM, Simon Pepping >>> wrote: > On Tue, Feb 22, 2011 at 11:25:20AM -0700, Glenn Adams wrote: >> I notice also that the nightly build target does not run all the >>> junit >> tests. It would be better if it run all of them plus checkstyle and >> findbugs. > > Many junit tests require a display. Nightly builds are run in a > headless configuration, hence I had to disable many junit tests. At > nightly builds there is no one to check checkstyle and findbugs errors > and warnings; therefore there is no point in running them. > > Simon > >>> >> >
Re: junit tests in nightly builds
On 23/02/11 14:42, Glenn Adams wrote: > I guess we disagree, since I believe application quality and code quality > are related. And, further, I believe findbugs at least can identify real, > functional bugs (as opposed to checkstyle). While I agree with the above, I’m with Simon on this. Tests should be done within a continuous integration tool; Nightly builds serve a different purpose. I increasingly feel the need to set up continuous integration for the FOP project. The ASF provides several CI environments (Hudson, among others), at some point in the future I’m going to try them out and set up something. Hopefully sooner rather than later. Vincent > G. > > On Wed, Feb 23, 2011 at 2:01 AM, Simon Pepping wrote: > >> On Wed, Feb 23, 2011 at 01:15:34AM -0700, Glenn Adams wrote: >>> OK, understand on the junit headless issue. For checkstyle/findbugs it >> would >>> be useful to fail the nightly build if they do not pass. I will >> investigate >>> the necessary changes to enable this option, which I hope can be adopted. >> >> I would not agree. Nightly builds are a courtesy to the user. It would >> be good if we could guarantee that the builds pass the junit tests. >> But it is not relevant to the user whether they pass checkstyle and >> findbugs rules. These tests address the issue of code quality, not of >> application quality. >> >> Simon >> >>> On Wed, Feb 23, 2011 at 12:55 AM, Simon Pepping >> wrote: >>> On Tue, Feb 22, 2011 at 11:25:20AM -0700, Glenn Adams wrote: > I notice also that the nightly build target does not run all the >> junit > tests. It would be better if it run all of them plus checkstyle and > findbugs. Many junit tests require a display. Nightly builds are run in a headless configuration, hence I had to disable many junit tests. At nightly builds there is no one to check checkstyle and findbugs errors and warnings; therefore there is no point in running them. Simon >> >
Re: junit tests in nightly builds
I guess we disagree, since I believe application quality and code quality are related. And, further, I believe findbugs at least can identify real, functional bugs (as opposed to checkstyle). G. On Wed, Feb 23, 2011 at 2:01 AM, Simon Pepping wrote: > On Wed, Feb 23, 2011 at 01:15:34AM -0700, Glenn Adams wrote: > > OK, understand on the junit headless issue. For checkstyle/findbugs it > would > > be useful to fail the nightly build if they do not pass. I will > investigate > > the necessary changes to enable this option, which I hope can be adopted. > > I would not agree. Nightly builds are a courtesy to the user. It would > be good if we could guarantee that the builds pass the junit tests. > But it is not relevant to the user whether they pass checkstyle and > findbugs rules. These tests address the issue of code quality, not of > application quality. > > Simon > > > On Wed, Feb 23, 2011 at 12:55 AM, Simon Pepping >wrote: > > > > > On Tue, Feb 22, 2011 at 11:25:20AM -0700, Glenn Adams wrote: > > > > I notice also that the nightly build target does not run all the > junit > > > > tests. It would be better if it run all of them plus checkstyle and > > > > findbugs. > > > > > > Many junit tests require a display. Nightly builds are run in a > > > headless configuration, hence I had to disable many junit tests. At > > > nightly builds there is no one to check checkstyle and findbugs errors > > > and warnings; therefore there is no point in running them. > > > > > > Simon > > > >
Re: junit tests in nightly builds
On Wed, Feb 23, 2011 at 01:15:34AM -0700, Glenn Adams wrote: > OK, understand on the junit headless issue. For checkstyle/findbugs it would > be useful to fail the nightly build if they do not pass. I will investigate > the necessary changes to enable this option, which I hope can be adopted. I would not agree. Nightly builds are a courtesy to the user. It would be good if we could guarantee that the builds pass the junit tests. But it is not relevant to the user whether they pass checkstyle and findbugs rules. These tests address the issue of code quality, not of application quality. Simon > On Wed, Feb 23, 2011 at 12:55 AM, Simon Pepping wrote: > > > On Tue, Feb 22, 2011 at 11:25:20AM -0700, Glenn Adams wrote: > > > I notice also that the nightly build target does not run all the junit > > > tests. It would be better if it run all of them plus checkstyle and > > > findbugs. > > > > Many junit tests require a display. Nightly builds are run in a > > headless configuration, hence I had to disable many junit tests. At > > nightly builds there is no one to check checkstyle and findbugs errors > > and warnings; therefore there is no point in running them. > > > > Simon > >
Re: junit tests in nightly builds [was: Re: Solving FindBugs issue]
BTW, if someone would bother to nominate me as a committer, I would be happy to do this work directly. I don't like depending on the other committers to make progress in this area any more than you like having to address it. G. On Wed, Feb 23, 2011 at 1:15 AM, Glenn Adams wrote: > OK, understand on the junit headless issue. For checkstyle/findbugs it > would be useful to fail the nightly build if they do not pass. I will > investigate the necessary changes to enable this option, which I hope can be > adopted. > > G. > > > On Wed, Feb 23, 2011 at 12:55 AM, Simon Pepping wrote: > >> On Tue, Feb 22, 2011 at 11:25:20AM -0700, Glenn Adams wrote: >> > I notice also that the nightly build target does not run all the junit >> > tests. It would be better if it run all of them plus checkstyle and >> > findbugs. >> >> Many junit tests require a display. Nightly builds are run in a >> headless configuration, hence I had to disable many junit tests. At >> nightly builds there is no one to check checkstyle and findbugs errors >> and warnings; therefore there is no point in running them. >> >> Simon >> > >
Re: junit tests in nightly builds [was: Re: Solving FindBugs issue]
OK, understand on the junit headless issue. For checkstyle/findbugs it would be useful to fail the nightly build if they do not pass. I will investigate the necessary changes to enable this option, which I hope can be adopted. G. On Wed, Feb 23, 2011 at 12:55 AM, Simon Pepping wrote: > On Tue, Feb 22, 2011 at 11:25:20AM -0700, Glenn Adams wrote: > > I notice also that the nightly build target does not run all the junit > > tests. It would be better if it run all of them plus checkstyle and > > findbugs. > > Many junit tests require a display. Nightly builds are run in a > headless configuration, hence I had to disable many junit tests. At > nightly builds there is no one to check checkstyle and findbugs errors > and warnings; therefore there is no point in running them. > > Simon >
junit tests in nightly builds [was: Re: Solving FindBugs issue]
On Tue, Feb 22, 2011 at 11:25:20AM -0700, Glenn Adams wrote: > I notice also that the nightly build target does not run all the junit > tests. It would be better if it run all of them plus checkstyle and > findbugs. Many junit tests require a display. Nightly builds are run in a headless configuration, hence I had to disable many junit tests. At nightly builds there is no one to check checkstyle and findbugs errors and warnings; therefore there is no point in running them. Simon
