V3 testsuites

Benjamin Kosnik bkoz@redhat.com
Tue Nov 14 09:06:00 GMT 2000


> If memory serves, there were some problems with doing what we wanted to
> do using portable plain /bin/sh.  This assumption is probably killing
> all the non-GNU-ish people doing a 'make check' at the top-level dir.
> Was mkcheck supposed to survive this long?  :-)

Bash was picked because none of us felt like doing the contortions 
necessary to get plain /bin/sh to work.

No, mkcheck wasn't supposed to be alive at this point. It was supposed to 
be a quick hack so that I could get a decent idea about

1) changes to the headers making compile times larger
2) changes that made the binary sizes of executables bigger.

I couldn't figure out a way to do this with dejagnu, but perhaps I was 
just being dense.

> Maybe 'make check' should just do what we do now in mkcheck for DejaGNU
> (echo some things and exit for now) without actually running mkcheck, and
> only 'make check-install' or another target entirely should run mkcheck...

...sounds good, assuming dejagnu can't do 1 and 2 above. (We could make 
the current Makefile rule for 'make check' turn into something like 
'make check-performance-analysis' or whatever.) Let's make the 
default be dejagnu so that everybody who doesn't care about the other 
stuff can quickly see if their libraries are ok. 

thanks for doing this gaby.

-benjamin


More information about the Libstdc++ mailing list