If you try building createrepo_c (my build is in doit.sh), you will see
…
RPM5 builds against libgit2 tip and libgit2 “breaks” every other month.
The fixes are usually rather easy, but the code is entirely proof-of-concept.
I recommend
rm -rf libgit2
(which is likely enuf to disable) and building —without-libgit2
…
Digging into the python test failures in build/Testing/Temporary/LastTest.log,
the (first) symptom is
test_contentstat (tests.test_contentstat.TestCaseContentStat) ... error:
(null): Signature: sigh tag[2]: BAD, tag 269 type 6 offset 20 count 1
(process:128): C_CREATEREPOLIB-WARNING **: read_header: rpmReadPackageFile()
error
ERROR
This is MANDATORY signature checking in RPM5: the test packages
tests/testdata/packages/*.rpm
are (likely, not checked) unsigned.
A similar failure is going to be seen everywhere that rpmReaqdPackageFile()
is called with unsigned packages built by rpm.org (rpm5 packages are _ALWAYS_
signed).
You can still disable MANDATORY signature checking by changing a #define in
system.h
...
/**
* Eliminate signature/digest disablers.
*/
#define SUPPORT_NOSIGNATURES 1
#define SUPPORT_NODIGESTS 1
Please note that rpm-5.4.18 will be released with MANDATORY
signatures and all the code marked with those #defines removed
from the release.
==> YOU HAVE BEEN WARNED <==
(apologies for caps: I have been attempting MANDATORY signature
checking since 2004, and enduring conversations about —nosignatures
needs for almost 5 years now, its all rather frustrating).
The recommended fix (when rpm-5.4.18 is released) for reading unsigned packages
produced by rpm.org will be to sign the packages and import the pubkey. I
believe (but not yet checked) signing would fix the createrepo_c test failure
as well.
hth
73 de Jeff
