From: Christophe R. <cs...@ca...> - 2013-01-22 15:57:19
|
Hi, I know, a weekday. Nevertheless. I would like to freeze for the 1.1.4 release, and I've been asked by a concerned interested party to pay particular attention to regressions this time (reasonable, given the pretty regressy state of 1.1.3). In particular, it would be good to fix not just regressions from 1.1.3 but also regressions, maybe on obscure platforms, since relatively recent releases. I know of <http://src.knowledgetools.de/boinkmarks/tests> as one handy resource to investigate regressions and also support on odd platforms. I think cl-test-grid probably also provides some useful information in this respect but I haven't been able to find a handy url to look at e.g. library build problems in sbcl releases tracked over time. If anyone could dig something like that up, or anything else, that would be helpful. Thanks, Christophe |
From: Paul N. <pn...@va...> - 2013-01-22 17:25:56
|
Are there platforms you are thinking of that need particular testing this release? Regards, Paul Nathan Sent from my iPhone On Jan 22, 2013, at 7:57 AM, Christophe Rhodes <cs...@ca...> wrote: > Hi, > > I know, a weekday. Nevertheless. I would like to freeze for the 1.1.4 > release, and I've been asked by a concerned interested party to pay > particular attention to regressions this time (reasonable, given the > pretty regressy state of 1.1.3). In particular, it would be good to fix > not just regressions from 1.1.3 but also regressions, maybe on obscure > platforms, since relatively recent releases. > > I know of <http://src.knowledgetools.de/boinkmarks/tests> as one handy > resource to investigate regressions and also support on odd platforms. > I think cl-test-grid probably also provides some useful information in > this respect but I haven't been able to find a handy url to look at > e.g. library build problems in sbcl releases tracked over time. If > anyone could dig something like that up, or anything else, that would be > helpful. > > Thanks, > > Christophe > > ------------------------------------------------------------------------------ > Master Visual Studio, SharePoint, SQL, ASP.NET, C# 2012, HTML5, CSS, > MVC, Windows 8 Apps, JavaScript and much more. Keep your skills current > with LearnDevNow - 3,200 step-by-step video tutorials by Microsoft > MVPs and experts. ON SALE this month only -- learn more at: > http://p.sf.net/sfu/learnnow-d2d > _______________________________________________ > Sbcl-devel mailing list > Sbc...@li... > https://lists.sourceforge.net/lists/listinfo/sbcl-devel > |
From: Christophe R. <cs...@ca...> - 2013-01-23 07:31:41
|
Paul Nathan <pn...@va...> writes: > Are there platforms you are thinking of that need particular testing this release? Well... Thanks to Christoph Egger, one particular one got tested: the Cheney collector has been broken (unbuildable) for a couple months. This affected primarily the mips platform (and alpha and hppa, but I think expecting those to still run is ambitious), and I've just pushed Christoph's fix. Other platforms I'm interested in include the BSDs -- are there failures not marked expected in the regression test suite? If there are, I'll take a patch marking them as expected, but on the other hand it would be good to fix the failure instead if possible. Odd build options? Is it still possible to build without :sb-eval? Without :sb-unicode? (Do we want to support those build options? I don't know, but I think if not we should aim to take them out explicitly). Cheers, Christophe |
From: Tomas H. <tom...@kn...> - 2013-01-22 17:44:15
|
Hi Christophe, On 01/22/2013 04:57 PM, Christophe Rhodes wrote: > I know of <http://src.knowledgetools.de/boinkmarks/tests> as one handy > resource to investigate regressions and also support on odd platforms. > I think cl-test-grid probably also provides some useful information in > this respect but I haven't been able to find a handy url to look at > e.g. library build problems in sbcl releases tracked over time. If > anyone could dig something like that up, or anything else, that would be > helpful. if you look at the table with test runs at <http://src.knowledgetools.de/boinkmarks/tests>, you can click on a row (needs javascript but I could be implemented without javascript). It will get selected. If you then click on another row, it will display diff between the test results for the two rows. Then there is the version table. For example, if you click on the 1.1.2.13-dea1e42 link, you'll see list of builds and test results on different machines for a particular version. There are some TODO links which I'd like to implement, but haven't got around to it yet. Those links will make the build and test logs available for download at some point. Maybe a simple button that lists regressions for the selected version could be added there? The biggest problem I have is that the builds break boinkmarks causing it to hang or breaking test result parsing (sometimes caused by sbcl changes, sometimes by cosmic rays...) which needs my manual intervention:-( I've added a feed <http://src.knowledgetools.de/boinkmarks/rss> as you wished for. It's a start but I am not entirely happy with the content. Unfortunatelly, the ui is pretty interactive without predictable urls. That could be fixed or improved but not in the near future. Cheers, Tomas |
From: Anton V. <avo...@ya...> - 2013-01-23 09:38:39
Attachments:
sbcl-reports.lisp
|
22.01.2013, 19:57, "Christophe Rhodes" <cs...@ca...>: > Hi, > > I know, a weekday. Nevertheless. I would like to freeze for the 1.1.4 > release, and I've been asked by a concerned interested party to pay > particular attention to regressions this time (reasonable, given the > pretty regressy state of 1.1.3). In particular, it would be good to fix > not just regressions from 1.1.3 but also regressions, maybe on obscure > platforms, since relatively recent releases. > > I know of <http://src.knowledgetools.de/boinkmarks/tests> as one handy > resource to investigate regressions and also support on odd platforms. > I think cl-test-grid probably also provides some useful information in > this respect but I haven't been able to find a handy url to look at > e.g. library build problems in sbcl releases tracked over time. If > anyone could dig something like that up, or anything else, that would be > helpful. boinkmarks is interesting project, I wish I've heard of it before, because it overlaps with cl-test-grid. As for cl-test-grid, it's test results are published online as lisp data. Assuming you or somebody else have run the tests, you can do (print-compiler-diff "sbcl/sbcl-diff.html" *all-results* "quicklisp 2012-12-23" "sbcl-1.1.0-linux-x86" "sbcl-1.1.3.12-b9691ef-linux-x86") And get an html table displaying results different between the two specified SBCL versions: http://common-lisp.net/project/cl-test-grid/sbcl/sbcl-diff.html As you can see, I've already tested sbcl-1.1.3.12-b9691ef-linux-x86 and there is no regressions comparing to sbcl-1.1.0, in the tests performed by cl-test-grid. Which are: - ql:quickoad every ASDF system in quicklisp - run testsuites of some the most often used libraries (57 libraries currently). Note, the several difference in results presented on the page are not improvements either. They are caused by problem in particular libraries such as elephant, which generates some my-config.sexp file during build, sometimes having this file unreadable. I suggest you to try generating cl-test-grid reports. You then be able to run tests and hunt for regressions yourself. Try the attached file from slime session. Best regards, - Anton |