SWI PSR corruption

[Christian Ludlam]

On 28 Apr Tom Hughes wrote:

> In message <9324bf624d.christian at venice.lab>
>         Christian Ludlam <chris at recoil.org> wrote:
> 
> > When you build an application targeting 32 bit OSs, it seems you MUST use
> > the -zpq512 compiler switch to tell the compiler that inline SWI calls
> > corrupt the PSR (why this isn't the default I don't know).
> 
> So there is a workaround then... John didn't mention that when I reported
> this bug a while ago.
> 
> > Of course, the -zpq command line isn't documented and is subject to
> > change. What's the recommended workaround?
> 
> File a bug with Castle like I did? That and add -Dswi to my command
> lines to stop SWI inlining.

OK. I was tempted to say add a release note saying something like "if you use
Norcroft versions nn to nn, put -zpq512 on the command line".

-- 
Christian Ludlam
chris at recoil.org