Re: [Tiger-devel] Notes on 3.2rc3

[Javier Fernández-Sanguino Peña]

On Sun, Apr 27, 2003 at 08:40:51AM -0700, Steve G wrote:
> Hello,
> 
> I've spent some more time with rc3 on a RH9 system and have
> some observations to share:
> 
> 1) In the Checking Cron entries I see this as the tests run
> on stdout/stderr:
> 11:23> Checking cron entries...
> /bin/ls: *: No such file or directory

Yes. I have this under control, it's due to the gen_cron script:
    (
      cd $dir
      $LS -l $LSGROUP *
    ) |
Since some of the directories might be empty the '*' character does not get 
expanded. I will fix it in the future.

> 
> 2) scripts/check_cron line 73 has LS twice.

Thanks for noticing, fixed.

> 
> 3) scripts/check_passwd, line 181. The ! is bad. Man page
> says pwck returns 0 on success. If it returns anything else
> it should have a warning.

Will check.

> 
> 4) `strings /usr/sbin/sendmail` produces nothing usable on
> a RH system. I would skip this test for RH systems or rely
> on the rpm reported version if the rpm package passes
> verification.

Yes. The check:
strings /usr/sbin/sendmail |grep '[0-9][0-9]/[0-9][0-9]/[0-9][0-9]'

probably is not reliable anymore. Many mailers might provide a 
/usr/sbin/sendmail which contains that string (exim, for example, does).
I will have to make this test depend on the existance of SENDMAIL-CF file 
and disable it in the tigerrc for Linux.

> 
> 5) Issue tests fail even though /etc/issue & /etc/issue.net
> exist. The haveallfiles command seems to be wrong or the
> ISSUEFILE variable isn't set. It should be in
> systems/Linux/2/config. BTW, that file says just issue,
> should it be a full path?

This is a _local_ check (probably should be disabled in the distribution) 
since it will compare /etc/issue & /etc/issue.net with ISSUEFILEs provided 
in the distribution. It's useful for organisations to check that systems 
provide the proper disclaimer. I will document it better and disable it 
from the default tigerrc.

> 
> 6) On the first pass, I had TigerNoBuild=Y in tigerrc. The
> embedded pathnames test did not complain that the REALPATH
> command didn't exist. It shouldn't have continued without
> that file.
> 

If realpath does not exist it will default to 'echo' which will work in 
most cases. 

> Overall, 3.2rc3 looks much better than 3.1.
> 
> Hope this helps,
> Steve Grubb

Helps a lot. I will try to fix these issues soon and provide a 3.2rc4 
version. I would still like some more comments/issues in more platforms so 
I will hold this until I get more feedback (or if I get none until two 
weeks time, probably)

Regards

Javi

pgpDAlPDy9gx6.pgp
Description: PGP signature