Обсуждение: New Linux/libc5 CVSup client
I've posted a new statically-linked CVSup client for linux-libc5 boxes in: ftp://postgresql.org/pub/CVSup/cvsup-15.4.2-client-linux-libc5.tar.gz I've noticed many checksum errors when using my older cvsup-15.2 installation, but have not yet had a real update to test this newer package to see if it helps. I used my existing from-sources Modula-3 installation to build it. That form of Modula-3 installation may not be for everyone, since it took over 200MB of disk space during the installation and ~50MB after. But... I found an rpm-packaged version of Modula-3 which was trivial to install on my glibc2 box, but have not yet been successful at building cvsup for that machine. It looks like a configuration problem of some sort, since it fails on the very first m3makefile statement. If anyone has an interest in it I can give them the urls and point them at the CVSup author; not sure if I'll have time to pursue it for a while... - Tom
Hi, I'm the author of CVSup. I'm really glad you cc'd me on your mail. On 02-Dec-98 Thomas G. Lockhart wrote: > I've posted a new statically-linked CVSup client for linux-libc5 boxes > in: > > ftp://postgresql.org/pub/CVSup/cvsup-15.4.2-client-linux-libc5.tar.gz Great! I've been meaning to try to get some newer binaries for platforms I don't have. > I've noticed many checksum errors when using my older cvsup-15.2 > installation, but have not yet had a real update to test this newer > package to see if it helps. It should fix that problem. The checksum errors were caused by some whitespace changes in RCS files that were introduced in recent versions of CVS. I made the checksumming algorithm for RCS files ignore inconsequential whitespace differences, so this problem shouldn't recur no matter what they do to CVS in the future. There's still one potential source of checksum errors. The very newest versions of CVS support some new features such as preserving file modes and handling links. These are implemented using RCS file extensions (see rcsfile(5) for details) which are not yet supported by CVSup. Luckily, it seems that few people use those extensions. I am going to implement a general scheme for handling them, hopefully for the next release of CVSup. Even if some files use the new extensions, the current version will update them OK. You'll see checksum errors, but the "fixups" at the end will correct them (slower, of course). > I used my existing from-sources Modula-3 installation to build it. That > form of Modula-3 installation may not be for everyone, since it took > over 200MB of disk space during the installation and ~50MB after. But... > > I found an rpm-packaged version of Modula-3 which was trivial to install > on my glibc2 box, but have not yet been successful at building cvsup for > that machine. It looks like a configuration problem of some sort, since > it fails on the very first m3makefile statement. This is the so-called "PM3" release of Modula-3, right? I've made the necessary changes to build CVSup under that version of Modula-3. I've been holding off on releasing it because I'd like to get some other new features into the program first. But it's well tested on a couple of FreeBSD mirror sites. I put the sources up for FTP at: ftp://ftp.polstra.com/private/lockhart/cvsup-15.5.tar.gz The MD5 checksum is 1f2466e83d06beb1ab2a2400515f9817. You should be able to build it with "make M3TARGET=LINUXLIBC6". I'd love to find out whether it works OK under Linux, and fix it if it doesn't. I'm going to be on vacation next week, so I may be a bit unresponsive to mail during that time. John --- John Polstra jdp@polstra.com John D. Polstra & Co., Inc. Seattle, Washington USA "Nobody ever went broke underestimating the taste of the American public." -- H. L. Mencken
> > I've noticed many checksum errors when using my older cvsup-15.2 > > installation, but have not yet had a real update to test this newer > > package to see if it helps. > It should fix that problem. The checksum errors were caused by > some whitespace changes in RCS files that were introduced in recent > versions of CVS. I made the checksumming algorithm for RCS files > ignore inconsequential whitespace differences, so this problem > shouldn't recur no matter what they do to CVS in the future. I wasn't complaining; it's nice that your fallback technique of full transfer made it a "don't care". > This is the so-called "PM3" release of Modula-3, right? Yes. Any opinions on it? Hopefully it isn't too incompatible with other releases. They seem to be enthusiastic, and RPM support really helps with the initial installation... > I'm going to be on vacation next week, so I may be a bit unresponsive > to mail during that time. No problem. My glibc2 machine is at work, so I'll test there tomorrow if I have time. Have a good vacation if I don't talk to you before. - Tom
On 03-Dec-98 Thomas G. Lockhart wrote: > I wasn't complaining; it's nice that your fallback technique of full > transfer made it a "don't care". Oh, I didn't think you were complaining; I was just explaining. I always worry that when people see those "checksum mismatch" messages, they'll think it's because the program is buggy. Then they might decide not to trust it at all. But because of the safeguards that it contains ("fixups"), it's impossible (to the best of my knowledge) for the package to deliver a corrupted file. Even if you kill it manually at any point, you'll either get a correct file or you'll still have your original version. >> This is the so-called "PM3" release of Modula-3, right? > > Yes. Any opinions on it? It's _the_ free release to use these days. (There's also a commercial M3 put out by Critical Mass <http://www.cmass.com/>.) DEC SRC hasn't done anything with their M3 release for ages now. Some of the stalwarts took over maintenance in their spare time, and the result is PM3. Most of the progress has been in the Linux arena, since most of the PM3 people run Linux. I've gotten it ported to FreeBSD, but haven't submitted the changes back to the PM3 folks yet. There are still some wrinkles to iron out, due to the fact that FreeBSD is in a transition period from a.out to ELF. Anyway, for Linux PM3 seems to be a big step forward. At least they are keeping up with the libc changes that keep coming along lately. > Hopefully it isn't too incompatible with other releases. It's mostly compatible except for some details of the build system that most programs never run into. > They seem to be enthusiastic, and RPM support really helps with the > initial installation... Yes indeed. > My glibc2 machine is at work, so I'll test there tomorrow if I have > time. Have a good vacation if I don't talk to you before. Thanks! John --- John Polstra jdp@polstra.com John D. Polstra & Co., Inc. Seattle, Washington USA "Nobody ever went broke underestimating the taste of the American public." -- H. L. Mencken