Hi,
Richard W.M. Jones wrote:
On Mon, Nov 09, 2020 at 11:17:38AM -0500, Todd Zullinger wrote:
The cvs and cvsps BR's are for the test suite, since we prefer to use the comprehensive test suite that git includes. So dropping those BR's is not a useful option.
The test suite still ran, and the subpackage still got built.
However in the test suite the cvsimport tests were skipped:
t9600-cvsimport.sh ................................. skipped: skipping cvsimport tests, cvs not found t9601-cvsimport-vendor-branch.sh ................... skipped: skipping cvsimport tests, cvs not found t9602-cvsimport-branches-tags.sh ................... skipped: skipping cvsimport tests, cvs not found t9603-cvsimport-patchsets.sh ....................... skipped: skipping cvsimport tests, cvs not found t9604-cvsimport-timestamps.sh ...................... skipped: skipping cvsimport tests, cvs not found
Yep, exactly. If we simply removed the BR's, we'd just be shipping a subpackage with commands that weren't tested and wouldn't work without cvs. Not a good user-experience. :)
I'll keep my eye on the progress of removing the xinetd dependency from cvs and will be sure to disable cvs in git if that doesn't happen for some reason. Hopefully it doesn't come to that.
In that case, a change like this should be all we need:
diff --git i/git.spec w/git.spec index 5188290..e85363b 100644 --- i/git.spec +++ w/git.spec @@ -66,8 +66,8 @@ %endif
# Allow cvs subpackage to be toggled via --with/--without -# Disable cvs subpackage by default on EL > 7 -%if 0%{?rhel} > 7 +# Disable cvs subpackage by default on Fedora >= 34 and EL > 7 +%if 0%{?fedora} >= 34 || 0%{?rhel} > 7 %bcond_with cvs %else %bcond_without cvs