On 20 January 2018 at 23:36, Juan Quintela <address@hidden> wrote:
Peter Maydell <address@hidden> wrote:
On 19 January 2018 at 16:43, Alexey Perevalov <address@hidden> wrote:
As I remember, I tested build in QEMU's docker build system,
but now I checked it on i386 Ubuntu, and yes linker says about unresolved
atomic symbols. Next week, I'll have a time to investigate it deeper.
This sounds like exactly the problem I pointed out in a previous
round of this patchset :-(
https://lists.gnu.org/archive/html/qemu-devel/2018-01/msg02103.html
Ignoring comments and sending patches anyway makes me grumpy,
especially when the result is exactly "fails obscurely on
some architectures only"...
It compiles for me. F25 i686 gcc. I did change it to use intptr_t
instead of uint64_t. So, I don't know what is going on here.
Did you change it to not use the 'nocheck' versions of the macros?
The code in master uses 'nocheck' which has exactly the effect
of masking this bug on i686...
So, I can agree that we have to fix anything that don't work, but I
can't agree that I didn't care about comments, at least I tried to fix
the problems you pointed me to.
I said the could should probably not use the nocheck macros. The
code in master is still using those macros, wrongly, which is why
this problem shows only on ppc32 and not all 32-bit hosts.