On 05-Mar-25 20:47, Ben Burton wrote: > > > If the test fails, the build should fail with an error message like > > "The installed libgcj-dev version does not match the installed gcj > > version." or similar. > > But this then _loses_ the guaranteed success of the build on the regular > architectures. Build-depends aren't just for autobuilders; they're for > humans also, and the package should build with any of the allowable > build-depends combinations. > > For your architecture the change would convert a guaranteed failure into > a possible failure, but for the release architectures it would convert a > guaranteed success into a possible failure. As a whole I'm not > comfortable with this.
You are right, 'libgcj4-dev | libgcj-dev' is a bad idea. The only thing which would be correct seems to be something like 'libgcj4-dev [!ppc64], libgcj6-dev [ppc64]'. This is also ugly and should not be used, because the ppc64 port is just a development project and not an official port at the moment. One solution which comes to my mind would be to make 'libgcj-dev' a real package like 'gcj' which is provided by 'gcc-defaults' and depends on the package from the current default gcc version. This would make a Build-Depends on just 'gcj, libgcj-dev' possible. Anyway, I do not see a reason to explicitly call 'gcj-3.3' instead of 'gcj' in the Makefiles. Regards Andreas Jochens -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
