Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 1 Jul 2019 16:09:52 -0400
From: Rich Felker <dalias@...c.org>
To: musl@...ts.openwall.com
Subject: Re: [PATCH] fix build failure on arm because of missing clz
 instruction

On Mon, Jul 01, 2019 at 12:09:26PM -0700, Andre McCurdy wrote:
> On Fri, Jun 28, 2019 at 9:19 PM Rich Felker <dalias@...c.org> wrote:
> > On Fri, Jun 28, 2019 at 03:55:56PM -0700, Andre McCurdy wrote:
> > > On Fri, Aug 24, 2018 at 4:20 PM Rich Felker <dalias@...c.org> wrote:
> > > > musl does not support being pure-thumb1 code, because some of the asm
> > > > source files are not thumb-compatible, but I think the C code can be
> > > > compiled as thumb1. -mthumb is only passed to the assembler for asm
> > > > source files if __thumb2__ is defined.
> > >
> > > Sorry to resurrect such an old thread, but it seems this patch was
> > > never applied?
> > >
> > > Without it, -mthumb -march=armv5t still fails to build due to clz
> > > getting into C code via inline assembler.
> >
> > Thanks for reviving this thread. I'll commit it or something similar.
> > I wonder if _ARM_ARCH>=5 && __thumb__!=1 would be a better test.
> 
> __thumb__ is either 1 (for both Thumb1 and Thumb2) or undefined (for
> ARM). The above might misleadingly suggest that it's sometimes defined
> with a value other than 1?

Oh, sorry. I meant _ARM_ARCH>=5 && (!__thumb__ || __thumb2__). I was
wrongly thinking __thumb__ was defined as 1/2 rather than having a
separate __thumb2__ macro.

Rich

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.