22 Jun
2017
22 Jun
'17
1:37 p.m.
On 6/22/17 9:06 AM, Thomas Petazzoni wrote:
Hello,
On Thu, 22 Jun 2017 08:47:06 -0400, Anthony G. Basile wrote:
Can you take a look at include/signal.h and tell me if knocks wants
__FAVOR_BSD defined or not? More generally, if _BSD_SOURCE and
_SVID_SOURCE are defined, do we want __FAVOR_BSD automatically defined
as well?
--
Anthony G. Basile, Ph.D.
Gentoo Linux Developer [Hardened]
E-Mail : blueness(a)gentoo.org
GnuPG FP : 1FED FAD9 D82C 52A5 3BAB DC79 9384 FA6E F52D 4BBA
GnuPG ID : F52D4BBA
Yes, I'm worried about regressions.
__FAVOR_BSD is still used in the
following:
include/setjmp.h
include/unistd.h
include/signal.h
include/features.h
include/netinet/udp.h
and deals with the differences in which signals and long jmps were
handled in older BSD. I don't know buildroot that well, but if you grep
the tree, will it show where __FAVOR_BSD is used? Or does it (like
gentoo's portage) pull in source code as it builds. I assume the latter
so grepping is probably insufficient.
Buildroot is a build system, it doesn't contain code. It downloads the
source code of the selected packages, as part of the build process.
In addition, the knock package is not using __FAVOR_BSD anywhere, so
this problematic package would not even match on a grep __FAVOR_BSD.
What knock package is doing is that it assumes a BSD-style definition
of tcphdr will be provided if _BSD_SOURCE and _SVID_SOURCE are defined.
Best regards,
Thomas