tmp-lshift.s:106: Error: selected processor does not support ARM mode `vdup.32 d6, r3'

Torbjörn Granlund tg at
Mon Dec 31 15:25:35 UTC 2018

Jeffrey Walton <noloader at> writes:

  No. Me overriding your flags is:

      ./configure ....
      make CFLAGS="..."

  In this case I stomp on your flags.

Now I get it.  You don't set CFLAGS, nor do you override it.  You stomp
on it.  Glad we have that crucial distinction clarified!

  > If you want your build to proceed, start without overriding CFLAGS.
  > Does that work?  Great, work from there.

  No, we cannot work from there. GMP selects screwed up release flags.
  They are dangerous and insecure.

Now, now!  How good your build failed then, keeping you out of immediate
danger.  ;-)

  GMP is the only project I know that cannot handle simple release build
  flags. And I build 66 of them here: .

You might want to take a look at GMP before you talk loudly about it.
We have lots of assembly code which need much more complex configuration
than your average Joe Package.

Surely GMP's configuration system can be improved in various ways.
Making it somehow handle contradictory tuples and CFLAGS is not an
improvement, though.  (What do you think GMP should do in your case?
Silently change configuration tuple?  Edit CFLAGS to match the tuple?
Do you really want that sort if resilience against user errors?)

  GMP really should follow GNU Coding Standards. It is not that hard.

You might want to read Niels' reply.  And RT Fine M might do you well

(This is the end of this dicussion for me.  Feel free to continue here
or elsewhere.)

Happy GNU Year!

Please encrypt, key id 0xC8601622

More information about the gmp-bugs mailing list