Unsubscribe > On Jan 9, 2017, at 1:18 AM, Helmut Grohne <hel...@subdivi.de> wrote: > > Source: tar > Version: 1.29b-1.1 > User: helm...@debian.org > Usertags: rebootstrap > > Hi, > > I am reporting this issue as a joint discovery with Roger Leigh > (X-Debbugs-CCed). > > While building tar, it detects the availability of xattrs/acls. The > relevant configure check resides in the macro gl_FILE_HAS_ACL which > lives in m4/acl.m4. The relevant test has the cache variable > gl_cv_getxattr_with_posix_acls. The test results are: > > build arch | host arch | result > -------------+--------------+------- > linux-any | linux-any | yes > kfreebsd-any | kfreebsd-any | no > linux-any | kfreebsd-any | yes > > The observed behaviour is easily explained by the fact that the compiled > program includes <linux/xattr.h>. Since linux headers are not installed > into multiarch paths yet, any compiler will find them as soon as > linux-libc-dev is installed. Thus this test effectively tests whether we > are building on linux or for linux. > > Roger Leigh pointed out that FreeBSD does support xattrs and acls, which > leads me to the conclusion that the test is operating wrongly. Maybe the > include directive could be made conditional to #ifdef __linux__? More > investigation is needed here, which is why I am Ccing d-bsd@l.d.o. > > This bug has blocked bootstrapping kfreebsd-any for a while now, because > I originally thought it was #837351 and nobody else had time to look > into it (not sure what happened to Steven). It'd help a lot with keeping > the bsd ports alive if FTBFS in essential packages could be fixed in a > more timely manner. I guess that's also the reason why it got kicked > from testing. :-/ > > In rebootstrap, I'll be working around this issue now and hand it over > to the bsd porters. > > Helmut >
signature.asc
Description: Message signed with OpenPGP
