Re: Clang now builds world and kernel, on i386 and amd64

From: Daniel Nebdal <dnebdal_at_gmail.com>
Date: Wed, 29 Sep 2010 10:25:58 +0200
On Wed, Sep 29, 2010 at 8:58 AM, Garrett Cooper <gcooper_at_freebsd.org> wrote:
> On Tue, Sep 28, 2010 at 11:43 PM, Dimitry Andric <dim_at_freebsd.org> wrote:
>> On 2010-09-29 02:28, Derek Tattersall wrote:
>>>
>>> A test shell script using mktemp (1) works fine on current built with
>>> clang today.  The clang case produces a filename with all "A"'s rather
>>> than the random letters expected.
>>
>> I cannot reproduce this on a system compiled entirely with clang:
>>
>> $ mktemp foo.XXXXXX
>> foo.MyUM5k
>> $ mktemp foo.XXXXXX
>> foo.YidMeT
>> $ mktemp foo.XXXXXX
>> foo.L27Cfz
>> $ mktemp foo.XXXXXX
>> foo.k3haLx
>>
>> ... and so on.  Can you post that test script, please?
>
> Please note your CPUTYPE and CFLAGS (for both those that had a problem
> and those that didn't) there might be some evidence in there that
> would help to resolve this issue with clang.
> Thanks,
> -Garrett


Works for me with random names; tested with the File::Temp script
posted earlier.

Amd64 on a Core2-family Xeon. In a make buildenv - environment I have
CPUTYPE='', and no CFLAGS set.
As for version, err. I csup-ed the code on Sep 24, and VERSION is
'FreeBSD 9.0-CURRENT amd64 900021' . How do you find the r-number
anyway?

I can grab today's version and see if it still works for me.


If it matters, the process was buildworld with gcc, installworld,
buildworld with clang, installworld.

--
Daniel Nebdal
Received on Wed Sep 29 2010 - 06:50:19 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:07 UTC