On Tue, May 17, 2011 at 12:57:31PM +0100, Dr. David Alan Gilbert wrote:
* Jiri Olsa (jolsa(a)redhat.com) wrote:
SNIP
OK, so below is a patch that:
a) Switches to using isprint() when deciding whether to print the character
ok
b) Sets the LC_ALL to C in the test script to make sure isprint
always does
the same thing
ok
c) Switches to using bash for test.sh
ok
d) Changes the char tests to only use + and - in the tests and
fixes up
the expected result.
I changed test_char functions to return single values, since I dont see
any reason to use the math in here
SNIP
> > Another suggestion is not to separate test_long into CPU
specific
> > sysdeps; there are only a few
> > combinations and they'll be shared among lots of CPUs - i686 and ARM
> > are the same for those tests.
>
> we could make one global being run for i686/arm and separate x86_64
> or something like that
So I haven't changed this (or for that reason enabled the ARM test) in this
patch because I thought I'd leave this one for you. My suggestion is to
make a sysdeps/long32.c and sysdeps/long64.c (it's probably slightly more subtle
than that, but I think that will do) and then choose which one to use.
I'm not sure here, I dont have all the info
Is ARM here completelly the same as i686? nothing surprising like the char signess
difference..?
Since we have arguments support only for x86/x86_64/arm, I guess
having long32/64 would be ok in case that arm and i686 totally match for long type
But I kind of incline to the specific arch sysdeps, it seems more general
to me in case other arch is added which would not fit here
thoughts? ;)
thanks,
jirka