Re: static link csh/tcsh dumps core

[christos%astron.com@localhost (Christos Zoulas)]

In article <20130311100110.GD2374%mail.duskware.de@localhost>,
Martin Husemann  <martin%duskware.de@localhost> wrote:
>On Mon, Mar 11, 2013 at 12:59:29AM +0100, Anders Hjalmarsson wrote:
>> I found this in section 2.3.1 of 
>> "Digital UNIX Calling Standard for Alpha Systems"
>> Part Number: AA-PY8AC-TE
>
>Very usefull, indeed!
>
>So, this makes it still sound like a compiler bug. Matt, do you want to
>bring this up upstream?
>
>I'll test a fix and will report back.

I think that it is not relevant if $29 is considered a scratch register
or not. If the compiler is using it in a function that calls setjmp()
and longjmp(), then there should be agreement between the compiler and
the library authors on who is responsible for saving and restoring it.
In this case, it seems just simpler to go the conservative way and fix
setjmp/longjump instead of messing with the compiler.

christos