From: Peter Korsgaard Date: Tue, 08 Dec 2015 21:51:53 +0100 Subject: [Buildroot] [PATCH] package/xfsprogs: disable crc32selftest In-Reply-To: <1449511236-21470-1-git-send-email-yann.morin.1998@free.fr> (Yann E. MORIN's message of "Mon, 7 Dec 2015 19:00:36 +0100") References: <1449511236-21470-1-git-send-email-yann.morin.1998@free.fr> Message-ID: <87a8pkmzo6.fsf@dell.be.48ers.dk> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net >>>>> "Yann" == Yann E MORIN writes: > xfsprogs generates a crc32 table, and wants to test it. So it builds a > native crc32selftest binary. So far, so good; except the host may not > have the required headers (uuid/uuid.h), and probably libs (like > libuuid). > And anyway, testing the table on the host is flawed, as it does not mean > it would be correct on the target (due to differences in bitness, > endianness, alignment...). > So we just disable the crc32selftest for Buildroot. > Fixes: > http://autobuild.buildroot.org/results/5d0/5d0ed24b4d368fed100bc3c4959520ebaa48e693/ > http://autobuild.buildroot.org/results/aef/aef223ee2646599eaab7c82f63ff5f19569d69a5/ > http://autobuild.buildroot.org/results/fb6/fb6679f3b1473169702c303ee5060abe7559b0b3/ > .... > Signed-off-by: "Yann E. MORIN" > Cc: Vicente Olivert Riera > --- > package/xfsprogs/0002-no-crc32-checks.patch | 32 +++++++++++++++++++++++++++++ > 1 file changed, 32 insertions(+) > create mode 100644 package/xfsprogs/0002-no-crc32-checks.patch > diff --git a/package/xfsprogs/0002-no-crc32-checks.patch b/package/xfsprogs/0002-no-crc32-checks.patch > new file mode 100644 > index 0000000..2cf6634 > --- /dev/null > +++ b/package/xfsprogs/0002-no-crc32-checks.patch > @@ -0,0 +1,32 @@ > +libxfs: do not try to run the crc32selftest > + > +Even though the crc32selftest is natively compiled (because it is to be > +executed), it fails in cross-compilation as the host may lack the > +required headers, like uuid/uuid.h (e.g. in a minimal environment). > + > +Moreover, running the crc32selftest natively is completely wrong, > +because it passing on the host does not mean it would still pass n the > +target (because endianness or bitness or alignment differences). > + > +So, just disable running the crc32selftest altogether. > + > +Note that there's a remaining bug-in-hiding, because the crc32 table > +generator is natively built, but with the target CLFAGS. s/CLFAGS/CFLAGS/ Committed with that fixed, thanks. -- Bye, Peter Korsgaard