make check gcc-3_0-branch libstdc++ behaviour

[Phil Edwards]

On Wed, Apr 11, 2001 at 11:13:25AM -0500, Gordon Sadler wrote:
> I thought? I saw a mention of libstdc++ testing moving to dejagnu? If
> this is true, do I need to activate it differently?

The move to dejagnu has only taken place on the trunk at present.
The branch still uses the libstdc++-v3 script.


> As of a few days ago, my test.log is having some odd output from
> checking libstdc++:
> 
> +       +       27_io/istream_extractor_arith.cc
> -r      /home/gbsadler/gcc-3.0/i686-pc-linux-gnu/libstdc++-v3/mkcheck: printf: /home/gbsadler/cvs/gcc-3.0/libstdc++-v3/testsuite/27_io/istream_extractor_char.cc:61:: invalid number
> -r      /home/gbsadler/gcc-3.0/i686-pc-linux-gnu/libstdc++-v3/mkcheck: printf: /home/gbsadler/cvs/gcc-3.0/libstdc++-v3/testsuite/27_io/istream_extractor_char.cc:61:: invalid number
> 27_io/istream_extractor_char.cc
> 
> 
> I'm used to + -b -r et.al. but the printf:...invalid number is all new
> here. Maybe the switch to dejagnu is affecting me?

The file giving the error is mkcheck itself, which is the script.
(There is a printf shell command; that's what's printing the complaints.)

mkcheck hasn't changed on the branch since mid-Feb.  If you're seeing errors
"as of a few days ago," that means that some other part of the testsuite
is failing, and when mkcheck tries to interpret the output as a number,
it fails.  (For example, if something is printing "Segmentation fault"
when it should be printing "42", then printf would have a hard time feeding
that to "%d\n", etc.)


Phil

-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.