SEO

Hi all,

I tried this on linux.redhat.rpm, but got no responses. I'm hoping
I'll have better luck here. It seems to me like a setup related
question, but if I've posted inappropriately to this group, please
forgive me. I'll probably try something a little more general like
comp.os.linux or comp.os.linux.misc if need be.

I'm having a little difficulty here. I've got a little software
bundle which includes a number of RPMs and some installation scripts.
In one of the first installation scripts to run, part of the process
is to do an ``rpm --checksig'' on the RPMs just as sort of a sanity
check. It isn't trying to check to make sure that the RPMs came
from a trusted source or anything like that, it's only doing it as
a checksum to make sure the files didn't get injured in some way
as they got passed around from hither to yon.

I recently started seeing some really disturbing behavior: I can
take an RPM, run a checksig against it, and it will tell me ``md5 NOT
OK,'' then if I simply run the same command again, without changing
anything (don't ask me what possessed me to try that, I don't know),
it will tell me that all is well with the world. This, of course,
breaks my installation process. The funny part is, if I simply
re-run the same install script a second time, it will succeed.

The installation script exits immediately if the checksig fails,
so the successful run that follows is not the result of something
that happens after the failed checksig. On the second pass through
the installation process, of course, it does everything exactly
the same way it did the first time, so that shouldn't matter.

The RPM in question is a home-grown RPM. This install script has
been run certainly thousands of times by various folks, and all of
the sudden it starts doing this out of nowhere. I can make this
happen once in a blue moon running ``rpm --checksig'' against my
RPM manually on the command line, but it happens with a little
more regularity from the install script.

Has anyone else out there seen such weirdness? Any clues?

Thanks,
tg.

--
To reply by mail, remove all lower case letters in my return address
(tGARDNER@ElectEngrngCompSci.CaseWesResUniv.EDU).