On 7/21/19 4:53 PM, Jean-Marc Pigeon via blfs-dev wrote:
On 07/21/2019 04:04 PM, Bruce Dubbs via blfs-dev wrote:
On 7/21/19 2:37 PM, Jean-Marc Pigeon via blfs-dev wrote:
On 07/21/2019 01:46 PM, Bruce Dubbs via blfs-dev wrote:
On 7/20/19 9:21 PM, Ken Moffat via blfs-dev wrote:
On Sun, Jul 21, 2019 at 11:29:51AM +1000, Wayne Blaszczyk wrote:
On Sat, 2019-07-20 at 18:02 +0100, Ken Moffat via blfs-dev wrote:
On Sat, Jul 20, 2019 at 09:01:00AM -0400, Jean-Marc Pigeon via
blfs-dev wrote:
Solution is to add
#include <linux/sockios.h>
within
qt-everywhere-src-5.13.0/qtserialbus/src/plugins/canbus/socketcan/socketcanbackend.cpp
Hoping this help.
(I guess the qt-5.13.0-upstream_fixes-1.patch, need
to be slightly updated :) ).
Sounds as if this is more widespread :-(
https://lore.kernel.org/lkml/20190720174844.4b989d34@sf/T/#u
At a minimum, firefox, qemu, and something called linux-atm.
The problem appears to be caused by a change in a linux header. I am
reluctant to include a long header for what is basically a
devinition of
a constant.
For a temporary workaround in qt/qtwebengine we can use:
sed -i 's/SIOCGSTAMP/0x8906/' \
qtserialbus/src/plugins/canbus/socketcan/socketcanbackend.cpp \
qtwebengine/src/3rdparty/chromium/net/base/network_interfaces_linux.cc
I don't really like this, but I like including a linux/ header even
less. This is currently being tested, so I'm not 100% sure it will
work.
We can probably do the same for firefox and qemu. I don't know about
thunderbird or other packages. It is unfortunate that these required
changes all affect very large packages.
-- Bruce
My 2 cents:
Setting SIOCGSTAMP/0x8906 is "pure evil", I'll explain why later bellow.
First: Let see the QT problem from the layering stand point (My mantra
:) )
Qt is a package high enough within the software tool-chain, it should be
OS unaware. OS dependencies should be hidden somewhere within a
QT include (lets say qtos.h, and missing definition detected within the
qtos.h) OR (better) use an Interface library to make QT working on
different OS by hiding the socket intrinsic.
My "fix" proposal was very short sighted, my only goal was
to have my (near) 600 packages to be recompiled "without flow"
within the 3 hours mark.
From BLSF stand point all modification/patch are small local
adjustment/trick to have it working, the way the upstream will
carry (or not) the modification, is the upstream privilege.
But.. BLSF, other mission, (beside building from scratch) could be
to provide feedback to upstream, and lets them know
they are not "OS Blind"... as soon OS Team is adjusting its
code definition (and this is the OS team privilege) QT is in trouble,
please work out your layering!.
Now.... SIOCGSTAMP/0x8906 adjustment proposal not only keep the
OS dependency within QT. but introduce hardware dependency too.
Which from the layering stand point is "pure evil" :) .
How is this workaround any worse than including linux/sockios.h ?
As opposed to linux/sockios.h ?
#if __BITS_PER_LONG == 64 || (defined(__x86_64__) && defined(__ILP32__))
/* on 64-bit and x32, avoid the ?: operator */
#define SIOCGSTAMP SIOCGSTAMP_OLD
#define SIOCGSTAMPNS SIOCGSTAMPNS_OLD
#else
Where SIOCGSTAMP_OLD is in asm-generic/sockios.h:
#define SIOCGSTAMP_OLD 0x8906
I suppose for the 5.2.x kernel we could do:
sed -e 's/SIOCGSTAMP/&_OLD/' ...
But that's still the equivalent of what I had above, but the C
preprocessor does it instead of us directly.
Because it is a Hexadecimal value coming from nowhere.
I'm sure that Qt and/or glibc will provide a proper change relatively
soon. As I said, the workaround is temporary.
I was NOT trying to be mean, I understood is was temporary.
I didn't take it that way. It's just a technical discussion and I took
it as a technical opinion.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
