13 ms·

Debian is switching back to GLIBC

179 points by tshepang 12y ago

vezzy-fnord 12y ago
Odd. I had always been convinced that they switched back to glibc again a while ago, but it seems they're beginning to do it only now. A lot of people are against forks that they deem frivolous or unnecessary, but I think it's one of the most pragmatic and important parts of free software. It's a failsafe, a Second Amendment, if you will.
opinionperson 12y ago
whats a debian?
- nemasu 12y ago
  I think it's a type of penguin?
JoshTriplett 12y ago
This seems like a very similar situation to egcs and gcc: a fork created because the original project proved too painful to work with, and folded back in after sorting out the project governance issues that caused the fork in the first place.
- plorkyeran 12y ago
  It's not quite as drastic: with egcs they basically just killed gcc and renamed egcs, while here it was more of a merger of the projects. Very similar, though.
  - gumby 12y ago
    They are in a funny way opposite (and I certainly know whereof I speak: look at who wrote this message [0]). It's hard to remember, but back in 1997 forking was considered really terrible and to be avoided at all costs. When we split off to form egcs the main resistance was to the idea that there would be a fork at all. Gcc, and a bunch of other programs, had a single person who controlled the "official" release. The fact that Cygnus maintained its own release tree for its customers (it was still free -- it simply wasn't identical to the mainline, and in fact advanced a lot faster) was the cause of much angst and even some mistrust. My reason for making this fork was because the mainline was so far behind us, and we had a commitment to folding all our changes back into the main line. Since the gcc maintainer was the bottleneck, we simply declared our tree a new fork with the support and participation of other major developers. We were trying to be the opposite of exclusivist. But even so it took me months of mailing, calling and negotiating. Many hours on the phone with rms who of course predicted the doom of free software if we went ahead. And we developed a steering committee, the first as far as I know for a free software project. In those two ways egcs was a watershed; plenty of developers thankfully now know no other way. So last year glibc went the opposite way: it decided it no longer needed a steering committee. Cool! [0] https://gcc.gnu.org/news/announcement.html https://gcc.gnu.org/news/announcement.html
    - gpvos 12y ago
      rms was (is?) probably still traumatized from the emacs/xemacs split, which is still there and, as far as I can see, does impede emacs development.
      - riffraff 12y ago
        as an outsider, how does the (x)emacs split impede development? It seems emacs has been steadily, if somewhat slowly, improving (new releases etc).
        gpvos 12y ago
        Well, it was a bit of a hunch, based on the slowness of development. Other commenters have maybe given better explanations.
      - e40 12y ago
        xemacs looks dead to me. 2009 was the last official release. See http://www.xemacs.org/Releases/index.html#Stable http://www.xemacs.org/Releases/index.html#Stable
      - hga 12y ago
        The former might be true, but it's very unlikely the latter, hard and harsh experience has shown that in the long term few things can impede development more than having RMS as a gatekeeper. He's a difficult person, or so I observed when we were in the same social circle, later when I was working for LMI while he was replicating the most essential new Zetalisp functionality, when we both had given up on that and were roommates when he launched the GNU project, and afterwords when for a while I worked for a "Software Hoarder", specifically the company that actually had a license for (James) Goslings Emacs, which GNU Emacs is an illicit fork of. (He even accused that company of setting fire to his apartment building; in truth it was pretty generic arson by "a couple of kids", not surprising because it was in a not so nice part of Cambridge.)
IvyMike 12y ago
I was unaware that drepper had left redhat (and glibc). He's now at Goldman Sachs. https://www.linkedin.com/in/ulrichdrepper https://www.linkedin.com/in/ulrichdrepper
- general_failure 12y ago
  Such great news. He was single handedly impeding the project.
  - jedisct1 12y ago
    So, does glibc provide strlcat() yet?
    - shmerl 12y ago
      I also wait for strlcpy. In cases when I need it on Linux I just take OpenBSD implementation: * http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/string/strlcpy.c?rev=1.11;content-type=text%2Fplain http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/string/st... * http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/string/strlcat.c?rev=1.13;content-type=text%2Fplain http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/string/st...
      - aktau 12y ago
        We use the linux kernel implementation at neovim. But a libc (and possibly asm optimized) implementation would be nice too (it could avoid two runs over the data). Not that it's likely to be a big improvement, strlen and memcpy are usually some of the most optimized functions there are.
        justincormack 12y ago
        In that case you need to change the license of neovim to GPL, as that is the license for Linux kernel code.
chengiz 12y ago
Some strong comments against Drepper. What's the back story here?
- damncabbage 12y ago
  He was the lead maintainer for glibc that people ended up needing to work around rather than with: https://sourceware.org/bugzilla/show_bug.cgi?id=4403 https://sourceware.org/bugzilla/show_bug.cgi?id=4403 https://sourceware.org/bugzilla/show_bug.cgi?id=5070 https://sourceware.org/bugzilla/show_bug.cgi?id=5070 https://sourceware.org/bugzilla/show_bug.cgi?id=5531 https://sourceware.org/bugzilla/show_bug.cgi?id=5531
  - kawsper 12y ago
    This one too: https://sourceware.org/bugzilla/show_bug.cgi?id=4980 https://sourceware.org/bugzilla/show_bug.cgi?id=4980
    - vezzy-fnord 12y ago
      And yet none of these hold a candle to this masterpiece: https://sourceware.org/bugzilla/show_bug.cgi?id=12701 https://sourceware.org/bugzilla/show_bug.cgi?id=12701 I dare you to top that.
      - haberman 12y ago
        These are so fantastic, keep them coming! I'm a little bit proud that I one time he told me "That's a good catch" (after verbally abusing someone else previously on the same bug). https://sourceware.org/bugzilla/show_bug.cgi?id=386 https://sourceware.org/bugzilla/show_bug.cgi?id=386
        rodgerd 12y ago
        I wish I could find his rant about his refusal to add binary coded decimals to the C standard, his refusal to allow anyone to add them to glibc, and his assertion that BCD is useless. Alas, it appears un-googlable.
mrmondo 12y ago
Anyone got another link? I'm getting "Error establishing a database connection"
- sharth 12y ago
  Here's a screengrab: http://i.imgur.com/XUNBIh1.png http://i.imgur.com/XUNBIh1.png
cryptonector 12y ago
Drepper's is a caustic personality. IIRC he's the reason that Linux has pre-linking instead of Solaris'direct binding, and so on.
- malkia 12y ago
  Can you share some details about this - like pros & cons once vs. another?
chrismonsanto 12y ago
Cache: http://webcache.googleusercontent.com/search?q=cache:VrGJCG7MXq4J:blog.aurel32.net/175&client=ubuntu&hl=en&gl=us&strip=1 http://webcache.googleusercontent.com/search?q=cache:VrGJCG7...
captainmuon 12y ago
Wow, Debian (and by extension Ubuntu) has been using eglibc for five years and I haven't noticed? I even compiled new versions of gcc + glibc (and installed them into $HOME, to run new binaries on older installations). You think I should have noticed that there was no glibc on the host system to begin with. But actually, the only thing I noticed is that the glibc (and gcc) build system is a bit crazy, and that it definitely will benefit from a new project lead.
- tshepang 12y ago
  You would not notice because it's practically the same. All files have the same names (afaik) for example. They are compatible, and eglibc was like a drop-in replacement. I believe they share nearly all code.

179 points by tshepang 12y ago

Odd. I had always been convinced that they switched back to glibc again a while ago, but it seems they're beginning to do it only now. A lot of people are against forks that they deem frivolous or unnecessary, but I think it's one of the most pragmatic and important parts of free software. It's a failsafe, a Second Amendment, if you will.

opinionperson 12y ago

whats a debian?

nemasu 12y ago
I think it's a type of penguin?

JoshTriplett 12y ago

This seems like a very similar situation to egcs and gcc: a fork created because the original project proved too painful to work with, and folded back in after sorting out the project governance issues that caused the fork in the first place.

plorkyeran 12y ago
It's not quite as drastic: with egcs they basically just killed gcc and renamed egcs, while here it was more of a merger of the projects. Very similar, though.
- gumby 12y ago
  They are in a funny way opposite (and I certainly know whereof I speak: look at who wrote this message [0]). It's hard to remember, but back in 1997 forking was considered really terrible and to be avoided at all costs. When we split off to form egcs the main resistance was to the idea that there would be a fork at all. Gcc, and a bunch of other programs, had a single person who controlled the "official" release. The fact that Cygnus maintained its own release tree for its customers (it was still free -- it simply wasn't identical to the mainline, and in fact advanced a lot faster) was the cause of much angst and even some mistrust. My reason for making this fork was because the mainline was so far behind us, and we had a commitment to folding all our changes back into the main line. Since the gcc maintainer was the bottleneck, we simply declared our tree a new fork with the support and participation of other major developers. We were trying to be the opposite of exclusivist. But even so it took me months of mailing, calling and negotiating. Many hours on the phone with rms who of course predicted the doom of free software if we went ahead. And we developed a steering committee, the first as far as I know for a free software project. In those two ways egcs was a watershed; plenty of developers thankfully now know no other way. So last year glibc went the opposite way: it decided it no longer needed a steering committee. Cool! [0] https://gcc.gnu.org/news/announcement.html https://gcc.gnu.org/news/announcement.html
  - gpvos 12y ago
    rms was (is?) probably still traumatized from the emacs/xemacs split, which is still there and, as far as I can see, does impede emacs development.
    - riffraff 12y ago
      as an outsider, how does the (x)emacs split impede development? It seems emacs has been steadily, if somewhat slowly, improving (new releases etc).
      - gpvos 12y ago
        Well, it was a bit of a hunch, based on the slowness of development. Other commenters have maybe given better explanations.
    - e40 12y ago
      xemacs looks dead to me. 2009 was the last official release. See http://www.xemacs.org/Releases/index.html#Stable http://www.xemacs.org/Releases/index.html#Stable
    - hga 12y ago
      The former might be true, but it's very unlikely the latter, hard and harsh experience has shown that in the long term few things can impede development more than having RMS as a gatekeeper. He's a difficult person, or so I observed when we were in the same social circle, later when I was working for LMI while he was replicating the most essential new Zetalisp functionality, when we both had given up on that and were roommates when he launched the GNU project, and afterwords when for a while I worked for a "Software Hoarder", specifically the company that actually had a license for (James) Goslings Emacs, which GNU Emacs is an illicit fork of. (He even accused that company of setting fire to his apartment building; in truth it was pretty generic arson by "a couple of kids", not surprising because it was in a not so nice part of Cambridge.)

IvyMike 12y ago

I was unaware that drepper had left redhat (and glibc). He's now at Goldman Sachs. https://www.linkedin.com/in/ulrichdrepper https://www.linkedin.com/in/ulrichdrepper

general_failure 12y ago
Such great news. He was single handedly impeding the project.
- jedisct1 12y ago
  So, does glibc provide strlcat() yet?
  - shmerl 12y ago
    I also wait for strlcpy. In cases when I need it on Linux I just take OpenBSD implementation: * http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/string/strlcpy.c?rev=1.11;content-type=text%2Fplain http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/string/st... * http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/string/strlcat.c?rev=1.13;content-type=text%2Fplain http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/string/st...
    - aktau 12y ago
      We use the linux kernel implementation at neovim. But a libc (and possibly asm optimized) implementation would be nice too (it could avoid two runs over the data). Not that it's likely to be a big improvement, strlen and memcpy are usually some of the most optimized functions there are.
      - justincormack 12y ago
        In that case you need to change the license of neovim to GPL, as that is the license for Linux kernel code.

chengiz 12y ago

Some strong comments against Drepper. What's the back story here?

damncabbage 12y ago
He was the lead maintainer for glibc that people ended up needing to work around rather than with: https://sourceware.org/bugzilla/show_bug.cgi?id=4403 https://sourceware.org/bugzilla/show_bug.cgi?id=4403 https://sourceware.org/bugzilla/show_bug.cgi?id=5070 https://sourceware.org/bugzilla/show_bug.cgi?id=5070 https://sourceware.org/bugzilla/show_bug.cgi?id=5531 https://sourceware.org/bugzilla/show_bug.cgi?id=5531
- kawsper 12y ago
  This one too: https://sourceware.org/bugzilla/show_bug.cgi?id=4980 https://sourceware.org/bugzilla/show_bug.cgi?id=4980
  - vezzy-fnord 12y ago
    And yet none of these hold a candle to this masterpiece: https://sourceware.org/bugzilla/show_bug.cgi?id=12701 https://sourceware.org/bugzilla/show_bug.cgi?id=12701 I dare you to top that.
    - haberman 12y ago
      These are so fantastic, keep them coming! I'm a little bit proud that I one time he told me "That's a good catch" (after verbally abusing someone else previously on the same bug). https://sourceware.org/bugzilla/show_bug.cgi?id=386 https://sourceware.org/bugzilla/show_bug.cgi?id=386
      - rodgerd 12y ago
        I wish I could find his rant about his refusal to add binary coded decimals to the C standard, his refusal to allow anyone to add them to glibc, and his assertion that BCD is useless. Alas, it appears un-googlable.

mrmondo 12y ago

Anyone got another link? I'm getting "Error establishing a database connection"

sharth 12y ago
Here's a screengrab: http://i.imgur.com/XUNBIh1.png http://i.imgur.com/XUNBIh1.png

cryptonector 12y ago

Drepper's is a caustic personality. IIRC he's the reason that Linux has pre-linking instead of Solaris'direct binding, and so on.

malkia 12y ago
Can you share some details about this - like pros & cons once vs. another?

chrismonsanto 12y ago

Cache: http://webcache.googleusercontent.com/search?q=cache:VrGJCG7MXq4J:blog.aurel32.net/175&client=ubuntu&hl=en&gl=us&strip=1 http://webcache.googleusercontent.com/search?q=cache:VrGJCG7...

captainmuon 12y ago

Wow, Debian (and by extension Ubuntu) has been using eglibc for five years and I haven't noticed? I even compiled new versions of gcc + glibc (and installed them into $HOME, to run new binaries on older installations). You think I should have noticed that there was no glibc on the host system to begin with. But actually, the only thing I noticed is that the glibc (and gcc) build system is a bit crazy, and that it definitely will benefit from a new project lead.

tshepang 12y ago
You would not notice because it's practically the same. All files have the same names (afaik) for example. They are compatible, and eglibc was like a drop-in replacement. I believe they share nearly all code.