Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
Hi Grant, Rémi and Yuri,
On 25-04-2007 20:30:45 -0500, Yuri Vasilevski wrote:
On Wed, 25 Apr 2007 23:39:47 +0200
Rémi Cardona [EMAIL PROTECTED] wrote:
Grant Goodyear a écrit :
Fabian Groffen wrote: [Sat Apr 14 2007, 03:33:03AM CDT]
For people that like reading it in html or via the web:
http://dev.gentoo.org/~grobian/gleps/glep-keywords.html
http://dev.gentoo.org/~grobian/gleps/glep-keywords.txt
So what would a version of Gentoo for amd64 based on FreeBSD but
using glibc be called? (It's not an entirely academic question;
Debian folks have worked on such a distribution for some time.)
I can't really tell from the text in your proposed GLEP.
I'm sure this GLEP can be extended later on should anyone feel like
doing a glibc-based freebsd port of gentoo (hurts my brains just
writing this :) )
I think it will be better if this scheme is specified in friendlier
way for future expansions, hence I this it should be more flexible.
I would propose this two modifications:
[snip]
So to give more examples,
A package that can only be build on arm, sparc and x86 with linux and
glibc or arm with uclibc can be specified as:
{arm,sparc,x86}:linux:glibc arm:uclibc
A package (lets say linux-headers) that makes sense on all systems that
support linux and only them can be specified as:
linux
[snip]
While I agree that you could be much more explicit in addressing the
exact thing that you're dealing with, I chose not to. The rationale
here is that the added complexity, as well as the added fine-grained
granularity is not necessary for at least now and what I would expect
from the reasonable future.
So in Grant's case, I would like to highlight that the right-hand field
of the keyword is an OS thing, not a kernel nor a userland. The reason
for this, is that it allows some freedom in what you consider to be OS
X (Not the Macintosh thing here). I'm not too familiar with FreeBSD and
it's flavours, so I'll talk about Solaris here. The SunOS kernel has
these days a few incarnations. OpenSolaris, Solaris, Nexenta...
For what we do, it seems that even though Nexenta has a GNU-based
userland, we can still address it in the same way as we can do for
Solaris, hence we don't need something special there. If we would, we
could make a keyword. Often times the setting of ELIBC and KERNEL helps
us to make the real decision where we need it (e.g. virtual/libintl),
which is set in the profiles, unrelated to the keyword.
--
Fabian Groffen
Gentoo on a different level
--
[EMAIL PROTECTED] mailing list
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Wed, 2007-04-25 at 09:35 +0200, Fabian Groffen wrote: Hereby I would like to request the counsel to discuss this mini-GLEP in the first meeting for which this request is in time. You got this in just in time for the next Council meeting. -- Chris Gianelloni Release Engineering Strategic Lead Alpha/AMD64/x86 Architecture Teams Games Developer/Council Member/Foundation Trustee Gentoo Foundation signature.asc Description: This is a digitally signed message part
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
Fabian Groffen wrote: [Sat Apr 14 2007, 03:33:03AM CDT] For people that like reading it in html or via the web: http://dev.gentoo.org/~grobian/gleps/glep-keywords.html http://dev.gentoo.org/~grobian/gleps/glep-keywords.txt So what would a version of Gentoo for amd64 based on FreeBSD but using glibc be called? (It's not an entirely academic question; Debian folks have worked on such a distribution for some time.) I can't really tell from the text in your proposed GLEP. -g2boojum- -- Grant Goodyear Gentoo Developer [EMAIL PROTECTED] http://www.gentoo.org/~g2boojum GPG Fingerprint: D706 9802 1663 DEF5 81B0 9573 A6DC 7152 E0F6 5B76 pgp10ctELm7SK.pgp Description: PGP signature
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
Grant Goodyear a écrit : Fabian Groffen wrote: [Sat Apr 14 2007, 03:33:03AM CDT] For people that like reading it in html or via the web: http://dev.gentoo.org/~grobian/gleps/glep-keywords.html http://dev.gentoo.org/~grobian/gleps/glep-keywords.txt So what would a version of Gentoo for amd64 based on FreeBSD but using glibc be called? (It's not an entirely academic question; Debian folks have worked on such a distribution for some time.) I can't really tell from the text in your proposed GLEP. I'm sure this GLEP can be extended later on should anyone feel like doing a glibc-based freebsd port of gentoo (hurts my brains just writing this :) ) Rémi -- [EMAIL PROTECTED] mailing list
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Wed, 25 Apr 2007 23:39:47 +0200
Rémi Cardona [EMAIL PROTECTED] wrote:
Grant Goodyear a écrit :
Fabian Groffen wrote: [Sat Apr 14 2007, 03:33:03AM CDT]
For people that like reading it in html or via the web:
http://dev.gentoo.org/~grobian/gleps/glep-keywords.html
http://dev.gentoo.org/~grobian/gleps/glep-keywords.txt
So what would a version of Gentoo for amd64 based on FreeBSD but
using glibc be called? (It's not an entirely academic question;
Debian folks have worked on such a distribution for some time.)
I can't really tell from the text in your proposed GLEP.
I'm sure this GLEP can be extended later on should anyone feel like
doing a glibc-based freebsd port of gentoo (hurts my brains just
writing this :) )
I think it will be better if this scheme is specified in friendlier
way for future expansions, hence I this it should be more flexible.
I would propose this two modifications:
(a) Instead of using n-tuples to describe a keyword, to use sets.
Both ways can be made semantically equivalent if we add the
restrictions that the sets of all the possible architecture, kernel,
userland, libc, ... sub-keywords are disjoint. (i.e. If there is a
userland sub-keyword bsd, then there can't be a kernel or libc
sub-keyword named bsd, they have to be named in a slightly different
way.)
But on the other hand, the notation will be way more flexible as a
keyword can only specify the relevant sub-keywords, while the rest can
be considered to be a wildcard (to mean all)
(b) In case there are several keywords A and B:
- if A is more specific that B, A takes precedence (ej. with !uclibc
arm:uclibc the package can be installed on any system that does not
uses uclibc as libc or on any system with arm hardware.)
- if A is exactly as specific as B, the union of A and B is used (ej.
!uclibc arm this is equivalent to the previous example.)
So to give more examples,
A package that can only be build on arm, sparc and x86 with linux and
glibc or arm with uclibc can be specified as:
{arm,sparc,x86}:linux:glibc arm:uclibc
A package (lets say linux-headers) that makes sense on all systems that
support linux and only them can be specified as:
linux
A package that is stable with gnu userland but still in testing with
bsd userland:
{some,arches}:gnu ~{some,arches}:bsd or just gnu ~bsd for all
arches.
Note that I propose to mark testing the whole set of sub-keywords, not
just like I run stable x86 with unstable bsd as I think it does not make
any sense as the resulting combination is still considered
unstable/testing.
There are two things I see against my proposal:
- This is not fully backward compatible, as it is currently equivalent
to normal arch keywords if the user runs linux with glibc/uclibc,
while it has a completely different meaning for Gentoo/Alt users.
But I don't see it as a big problem because, as far as I understand,
this will be just one of many changes that need to be made to make
Gentoo/Alt as integrated as GNU/Linux is now into portage.
- For this to work, the keyword resolver will have to be way more
complex than it is now, as it will have to compute the subset of all
possible keywords some ebuild defines to see if user's accept
keywords intersect it.
Kindest regards,
Yuri.
--
[EMAIL PROTECTED] mailing list
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Sat, 2007-04-14 at 05:41 -0700, Robin H. Johnson wrote: Anybody that feels like inspecting C code, look at mlmmj-1.2.14/src. getaddrsfromfd.c:27 - this mmap fails, 'Could not mmap fd: Invalid argument' mlmmj-send.c Is it possible we're hitting 1024 fd open? We have had a similar problem at my current employer. -- Chris Gianelloni Release Engineering Strategic Lead Alpha/AMD64/x86 Architecture Teams Games Developer/Council Member/Foundation Trustee Gentoo Foundation signature.asc Description: This is a digitally signed message part
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Mon, Apr 16, 2007 at 07:50:58AM -0400, Chris Gianelloni wrote: On Sat, 2007-04-14 at 05:41 -0700, Robin H. Johnson wrote: Anybody that feels like inspecting C code, look at mlmmj-1.2.14/src. getaddrsfromfd.c:27 - this mmap fails, 'Could not mmap fd: Invalid argument' mlmmj-send.c Is it possible we're hitting 1024 fd open? We have had a similar problem at my current employer. I raised all of the limits significantly in testing, with no effect. -- Robin Hugh Johnson Gentoo Linux Developer Council Member E-Mail : [EMAIL PROTECTED] GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85 pgpQfWCW9BI4U.pgp Description: PGP signature
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Tuesday 17 April 2007 01:10:20 Chris Gianelloni wrote: [ ... ] Yeah, ulimit won't do it. We hit this issue with mimedefang, actually. Our problem is that the kernel is doing the limiting. We ended up having to split things up a good bit into multiple processes to get everything working. We also added another machine to the cluster to try to reduce the load on any one server at a time. Nothing we did with ulimit made a bit of difference. http://www.cyberciti.biz/faq/linux-increase-the-maximum-number-of-open-files/ -- CS -- [EMAIL PROTECTED] mailing list
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Tue, Apr 17, 2007 at 11:19:24AM +1200, Christopher Sawtell wrote: Yeah, ulimit won't do it. We hit this issue with mimedefang, actually. Our problem is that the kernel is doing the limiting. We ended up having to split things up a good bit into multiple processes to get everything working. We also added another machine to the cluster to try to reduce the load on any one server at a time. Nothing we did with ulimit made a bit of difference. http://www.cyberciti.biz/faq/linux-increase-the-maximum-number-of-open-files/ # sysctl fs.file-max fs.file-max = 206524 And tracking fs.file-nr indicates that the total number of files open in the entire system barely exceeds 2000 when the lists are flying. -- Robin Hugh Johnson Gentoo Linux Developer Council Member E-Mail : [EMAIL PROTECTED] GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85 pgpnGVhvKhe2E.pgp Description: PGP signature
[gentoo-dev] Re: [GLEP] RFC - Keywording scheme
Fabian Groffen wrote: This GLEP has been laying around for some long time now in my gleps dir. I nearly forgot about it. Anyway, feedback is appreciated. Since it is a keywording scheme that is compatible with the scheme that is currently in use and fulfils all the requirements, it sounds like a no-brainer to me.. i'm sure someone will flame me for daring to say so but wtf ;) -- [EMAIL PROTECTED] mailing list
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Sat, Apr 14, 2007 at 07:32:12AM +0100, Steve Long wrote: Fabian Groffen wrote: This GLEP has been laying around for some long time now in my gleps dir. I nearly forgot about it. Anyway, feedback is appreciated. list-admin-hat Grobian: can you please resend your message to the list? This is the second message that I've seen lost in two days, and I'm annoyed now :-(. (This message is public so I can trace it as well). /list-admin-hat -- Robin Hugh Johnson Gentoo Linux Developer Council Member E-Mail : [EMAIL PROTECTED] GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85 pgpyFQZTEo7jy.pgp Description: PGP signature
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On 14-04-2007 01:19:41 -0700, Robin H. Johnson wrote: On Sat, Apr 14, 2007 at 07:32:12AM +0100, Steve Long wrote: Fabian Groffen wrote: This GLEP has been laying around for some long time now in my gleps dir. I nearly forgot about it. Anyway, feedback is appreciated. list-admin-hat Grobian: can you please resend your message to the list? Because I don't have it either, luckily there is GMane: http://thread.gmane.org/gmane.linux.gentoo.devel/48017 For ease of readers, some copy paste: For people that like reading it in html or via the web: http://dev.gentoo.org/~grobian/gleps/glep-keywords.html http://dev.gentoo.org/~grobian/gleps/glep-keywords.txt -- Fabian Groffen Gentoo on a different level GLEP: XXX2 Title: Keywording scheme Version: $Revision: 1.8 $ Last-Modified: $Date: 2007/04/13 17:26:35 $ Author: Diego Pettenò [EMAIL PROTECTED], Fabian Groffen [EMAIL PROTECTED] Status: Active Type: Standards Track Content-Type: text/x-rst Created: 11-Dec-2005 Post-History: 13-Apr-2007 Abstract This GLEP is a replacement of the keywording scheme from GLEP 22 [#GLEP22]_. The current use of keywords is retained in favour of 4-tuple keywords. This GLEP defines how current keywords are to be interpreted, and how future keywords should be constructed. Motivation == Although the state of GLEP 22 [#GLEP22]_ is final, its keywording scheme was never propagated through the tree. In fact, 4-tuple keywords are not used at all. This GLEP defines a keywording scheme that is compatible with the scheme that is currently in use. Rationale = The Gentoo/Alt project deals with different Operating Systems and architectures. Recently Gentoo/FreeBSD for Sparc was introduced after support for x86 platforms. This yielded in another new keyword. For these kind of platforms, a single field keyword is not enough to properly describe the OS and architecture. While four fields in a keyword are overkill, two fields in a keyword should be enough for everyone. Backwards Compatibility === The proposed keywording scheme is fully compatible with the current situation of the portage tree, this in contrast to GLEP 22. The variables provided by GLEP 22 can't be extracted from the new keyword, but since GLEP 22-style keywords aren't in the tree at the moment, that is not a problem. The same information can be extracted from the ``CHOST`` variable, if necessary. No modifications to ebuilds will have to be made. Specification = Keywords will consist out of two parts separated by a hyphen (``-``). The left hand part of the keyword is the architecture, such as `x86`, `sparc` or `ppc`. The right hand part indicates the operating system or distribution, such as `linux`, `macos`, `solaris` or `fbsd`. If the right hand part is omitted, it implies the operating system/distribution type is GNU/Linux. In such case the hyphen is also omitted. Examples of such keywords are ``x86`` and ``sparc-fbsd``. This is fully compatible with the current keywords used in the tree. Examples of OS/distributions for the right hand side of the keyword are: :: (linux) GNU/Linux (Gentoo biased, but not fixed) fbsd FreeBSD macosApple Mac OS solaris Sun Solaris Both architecture as well as OS/distribution are lower-case ASCII (alpha) numeric character sequences. A valid keyword matches the following expression: ``[a-z0-9]+(-[a-z0-9]+)?`` Note that no limit on the length of both fields in the keyword are imposed. However, we cannot overemphasize our preference to keep keywords small and sensible. .. [#GLEP22] GLEP 22, New keyword system to incorporate various userlands/kernels/archs, Goodyear, (http://glep.gentoo.org/glep-0022.html) Copyright = This document has been placed in the public domain.
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Saturday 14 April 2007, Fabian Groffen wrote: On 14-04-2007 01:19:41 -0700, Robin H. Johnson wrote: On Sat, Apr 14, 2007 at 07:32:12AM +0100, Steve Long wrote: Fabian Groffen wrote: This GLEP has been laying around for some long time now in my gleps dir. I nearly forgot about it. Anyway, feedback is appreciated. list-admin-hat Grobian: can you please resend your message to the list? Because I don't have it either, luckily there is GMane: http://thread.gmane.org/gmane.linux.gentoo.devel/48017 ive found hooking up my NNTP client to GMane and downloading missed e-mails from there works quite well if only i could say the same for our mailing list ;( -mike pgpwyH42OLDX7.pgp Description: PGP signature
Re: [gentoo-dev] Re: [GLEP] RFC - Keywording scheme
On Sat, Apr 14, 2007 at 07:47:12AM -0400, Mike Frysinger wrote: Because I don't have it either, luckily there is GMane: http://thread.gmane.org/gmane.linux.gentoo.devel/48017 ive found hooking up my NNTP client to GMane and downloading missed e-mails from there works quite well if only i could say the same for our mailing list ;( Unless Gmane has some deep foo in our servers that I don't know about, it's just as likely to be missing some list messages as our developers. The missing messages aren't in mlmmj's archive either. My rough stats show that we've had problems with ~1000 messages over the last 663 days - 1.5 messages per day. The core in this is a message not going to all recipients. As as I say, I have debug stuff on now, and it's given me one core dump that's pretty much garbage, and one other interesting bit of input. Anybody that feels like inspecting C code, look at mlmmj-1.2.14/src. getaddrsfromfd.c:27 - this mmap fails, 'Could not mmap fd: Invalid argument' mlmmj-send.c -- Robin Hugh Johnson Gentoo Linux Developer Council Member E-Mail : [EMAIL PROTECTED] GnuPG FP : 11AC BA4F 4778 E3F6 E4ED F38E B27B 944E 3488 4E85 pgpH9LLCzNjPd.pgp Description: PGP signature
