On Jun 2, 2007, at 1:45 AM, Michael Jennings wrote:
On Friday, 01 June 2007, at 10:30:04 (-0400),
Jeff Johnson wrote:
(aside) Yes there;'s breakage wrto zlib building rpm on x86_64. I
had noticed about 2 weeks ago (I don't normally have access to
x86_64) and the fix is in some tree of mine. cvs.rpm.org was
abandoned and rpm5.org was created in the last 2 weeks. Mea culpa
;-)
This sounds like the same problem I reported back in February.
Attached is the patch I used to fix it; is that what you were after,
or is this an unrelated issue?
Perhaps. The core issue is
How much of internal zlib needs to be maintained in rpm?
I've likely dropped much zlib baggage while setting up
the rpm-4_5 branch. I keep HEAD as close as possible
to upstream to simplify diff maintenance, but, well,
makevms.com ain't exactly useful or necessary to rpm-4_5.
I am still conflicted about whether minigzip should be
carried in /usr/lib/rpm. gzip and zlib differ, one cannot,
say, uncompress using zlib and recompress using gzip
and achieve bye-for-byte identical re[presentations. minigzip instead
of gzip does produce idempotent recompression.
The usage case for minigzip is whether scripting up
*.rpm package disassembly and reassembly across
rsync transport is worth "supporting". There is a large,
like 35->75 percent, bandwidth saving if one is willing
to slice-and-dice *.rpm packages. David Woodhouse
demonstrated that years ago with a kernel package.
OTOH, if you have to slice-and-dice and script for efficient
*.rpm transport, what good is a *.rpm package anyways?
I have similar mixed feelings wrto deltarpm transport.
Therein lies my conflict, whether to link minigzip and include
in /usr/lib/rpm or not, which is the root cause of the zlib
build failure.
Likely more than you wanted to know ;-)
73 de Jeff
______________________________________________________________________
RPM Package Manager http://rpm5.org
Developer Communication List rpm-devel@rpm5.org
