[BusyBox] Re: "rpm" fails with SIGSEGV

[Rob Landley]

On Wednesday 12 November 2003 04:56, Laurence Anderson wrote:
> >   so i renamed it simply to apache.rpm.  even with that name,
> > i could "rpm -qpi" and "rpm -qpl".  but when i tried to install, i
> > immediately got SIGSEGV.
>
> Just testing rpm for myself, the extraction code doesn't seem to work too
> well...
>
> First problem I have identified is that the new read_gz code won't
> necessarly return the requested amount of data; archive_xread_all &
> archive_xread_all_eof don't handle this, unlike bb_xread_all which does.
> Attched some kind of solution. This works a bit better, but fails after a
> few files complaining about invalid magic... however the cpio applet works
> fine (which uses exactly the same code)... maybe something todo with the
> file being compressed?

Huh.  I don't use RPM myself if I can avoid it, but...

Is this the new "fork and use a pipe" code?  (Glenn said he saw a bug like 
this.  I'm happy to try tracking it down if we can get a reproducible test 
case.  This is on a standard intel setup?  What kernel, what libc, etc...)

By the way, I've seen a pipe return 4k chunks (one page).  I don't know why, 
but I've seen this before.  And it doesn't seem to actually violate any spec, 
so you just stick a for loop in the wrapper.

archive_xread_all isn't the right wrapper though.  I'd fix the plain 
archive_xread to always return the full amount of data that was asked for 
even if the pipes are screwed up.  (A read of length 0 indicates EOF, but a 
short read does NOT indicate EOF.  Unix portability issue, I'm told.  Yes, I 
asked...)

Rob