[Mono-dev] mono 2.0, solaris x86, bootstrap

[Andreas Färber]

Am 05.10.2008 um 13:57 schrieb James Mansion:

> Andreas Färber wrote:
>> package. And most importantly many workarounds for Solaris 10 are not
>> necessary on OpenSolaris.
>>
> For any practical production use though, its necessary to support real
> Solaris 10, even if its
> only the latest version.

You've probably not followed the whole discussion. I have post-2.0  
Mono running on real Solaris 10 amd64, it works fine. Some others have  
reported class library bugs for Solaris 9 sparc, so it would in  
general seem to work there, too. And there used to be Solaris 8  
packages from Novell.

No one has been arguing for one or the other system here, the  
instructions are simply different for the two.


> Is there a list anywhere of the problematic bits of the system?

In the S10 resources pointed to recently, there's for instance...
- Sun's tar which doesn't extract long file names correctly (use GNU  
tar for tarballs)
- Sun's make (use GNU make instead, MAKE=gmake)
- Sun's C compiler (Mono requires gcc, CC=gcc etc.)
- Sun's mcs command (conflicts with Mono's mcs bootstrap when in $PATH)
- Solaris' GLib pkg-config files which are incompatible with gcc  
(replace -mt appropriately)
- pthreads was not correctly detected or passed on (specify --with- 
tls=pthread)

There also used to be some shell scripts incompatible with S10's  
Bourne shell (e.g., `export A=b` vs. `A=b; export A`). Hopefully  
that's fixed now.

> Lots of Solaris users are still in production on Solaris 8, too.  I'd
> dearly like to be able to
> consider Mono for real projects but that can't happen until it works
> properly on Sol8
> sparc and Sol10 (sparc and AMD).

Unfortunately, sparc is hard to test without access to the hardware.  
The only sparc emulator appeared to be QEMU, whose qemu-system-sparc  
is 32-bit only (sun4m), whereas Solaris 10 and Mono seem to expect  
sparcv9. I have a local branch where I once tried to make it work on a  
sparcv8 QEMU VM running Debian Etch, but the emulated compilation took  
much too long to get far.

As someone with access to the hardware, please tell us what the  
problem is on your Solaris 8 boxes. If you have submitted bug reports  
consider posting links here since the Solaris users among us are not  
automatically cc'ed in Bugzilla. Only few sparc users have been around  
the last few years, so waiting is unlikely to bring a solution. Is  
there a problem with Linear IR for instance or has that already been  
ported to sparc?

Andreas