Strange disappearance of test failures
Locked
Strange disappearance of test failures
 
|
I don't usually complain about tests passing, but in this case tests
have passed that used to fail, but not because of code changes. http://build.squeak.org/job/SqueakTrunk/272/testReport/ shows failures in decompiling and method highlighting that don't exist in http://build.squeak.org/job/SqueakTrunk/273/testReport/ The difference? I rebuilt the image under test: 273's image is 4.4-12327 updated to trunk, run through the release process [1] and then shrunk [2]. I _imagine_ the forced recompilation of everything is what's made the tests pass? frank [1] Form openAsBackground: (FileDirectory default / '..' / 'resources/wallpaper.png') pathName. Workspace allInstances do: [:w | w topView delete]. ReleaseBuilderFor4dot5 prepareNewBuild. "this calls Compiler recompileAll" Smalltalk snapshot: true andQuit: true. [2] #('39Deprecated' '311Deprecated' '45Deprecated' 'Nebraska' 'Universes' 'XML-Parser') do: [ :pkgName | (MCPackage named: pkgName) unload. MCMcmUpdater disableUpdatesOfPackage: pkgName]. Smalltalk cleanUp. |
Re: Strange disappearance of test failures
|
|
Yes, as I already said, I corrected the decompiler to be in line with the modification of Compiler by Eliot, and the recompilation of the system "fait monter la mayonnaise" (sorry, no time to dig for an english translation). 2013/4/17 Frank Shearar <[hidden email]> I don't usually complain about tests passing, but in this case tests |
Re: Strange disappearance of test failures
|
|
Yes, but that didn't happen in the last 24 hours did it? Because
that's the time difference between the two builds. Or do you mean that your decompilation fixes plus my Compiler recompileAll resulted in not creating "bad" stuff in the new updated-from-4.4 image? frank On 17 April 2013 13:57, Nicolas Cellier <[hidden email]> wrote: > Yes, as I already said, I corrected the decompiler to be in line with the > modification of Compiler by Eliot, and the recompilation of the system "fait > monter la mayonnaise" (sorry, no time to dig for an english translation). > > > 2013/4/17 Frank Shearar <[hidden email]> >> >> I don't usually complain about tests passing, but in this case tests >> have passed that used to fail, but not because of code changes. >> >> http://build.squeak.org/job/SqueakTrunk/272/testReport/ shows failures >> in decompiling and method highlighting that don't exist in >> http://build.squeak.org/job/SqueakTrunk/273/testReport/ >> >> The difference? I rebuilt the image under test: 273's image is >> 4.4-12327 updated to trunk, run through the release process [1] and >> then shrunk [2]. >> >> I _imagine_ the forced recompilation of everything is what's made the >> tests pass? >> >> frank >> >> [1] >> Form openAsBackground: (FileDirectory default / '..' / >> 'resources/wallpaper.png') pathName. >> Workspace allInstances do: [:w | w topView delete]. >> ReleaseBuilderFor4dot5 prepareNewBuild. "this calls Compiler recompileAll" >> Smalltalk snapshot: true andQuit: true. >> >> >> [2] >> #('39Deprecated' '311Deprecated' '45Deprecated' 'Nebraska' 'Universes' >> 'XML-Parser') do: [ :pkgName | >> (MCPackage named: pkgName) unload. >> MCMcmUpdater disableUpdatesOfPackage: pkgName]. >> >> Smalltalk cleanUp. >> > > > > |
Re: Strange disappearance of test failures
|
|
How can this happen ? Here is a theory (unverified, and with no code under my eyes): - we change Environment, this triggers a recompile of some methods, - then we fix Compiler/Decompiler but we do not trigger a recompile -> the image is in bad state It ill be solved - either by a new Environment change that would recompile the same methods - or a Compiler recompileAll
2013/4/17 Frank Shearar <[hidden email]> Yes, but that didn't happen in the last 24 hours did it? Because |
«
Return to Squeak - Dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |