Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 278 - Requirements for Gentoo v1.0
Summary: Requirements for Gentoo v1.0
Status: RESOLVED FIXED
Alias: None
Product: Gentoo Linux
Classification: Unclassified
Component: [OLD] Core system (show other bugs)
Hardware: x86 Linux
: High normal (vote)
Assignee: Daniel Robbins (RETIRED)
URL:
Whiteboard:
Keywords:
: 205 (view as bug list)
Depends on: 86 393 405 406 407 524
Blocks:
  Show dependency tree
 
Reported: 2002-01-22 17:34 UTC by Geert Bevin
Modified: 2003-02-04 19:42 UTC (History)
2 users (show)

See Also:
Package list:
Runtime testing required: ---


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Geert Bevin 2002-01-22 17:34:17 UTC
I think it's slowly getting time to think about the requirements for Gentoo
v1.0. Implementing those will make it possible to do a feature freeze and iron
out all outstanding bugs. Please add your comments / vision on what should and
shouldn't be there to make things advance.
Comment 1 Geert Bevin 2002-01-22 17:35:43 UTC
Portage 2.0 is imho a very important requirement. Portage is the center of the
entire distribution and it should be clean to start v1.0 with. A lot of little
userfriendlyness features should be added to smoother out the edges.
Comment 2 Mikael Hallendal (hallski) (RETIRED) gentoo-dev 2002-01-22 17:38:05 UTC
Agreed, it might also be time to start thinking of either freezing thee portage
tree and only allow security updates or updates for brooken packages or to
branch and start working on getting everything in the 1.0-branch to build perfectly.
Comment 3 Geert Bevin 2002-01-22 17:39:29 UTC
Deciding which packages will be inside 1.0 and which not is also important.
Updating old packages to the latest ebuild specs should also be done.
Comment 4 Mikael Hallendal (hallski) (RETIRED) gentoo-dev 2002-01-22 17:40:10 UTC
a descision on eclasses needs to be taken pretty soon.
Comment 5 Geert Bevin 2002-01-22 17:42:31 UTC
Developer and portage documentation should be updated.
Comment 6 Geert Bevin 2002-01-22 17:43:19 UTC
A document describing how and where Gentoo is different from other distribs
might be handy for people who are switching
Comment 7 Daniel Robbins (RETIRED) gentoo-dev 2002-01-24 10:05:42 UTC
We need a new smarter more convenient USE system and a rewrite of the downloader
in Portage ("resume" support)
Comment 8 Mikael Hallendal (hallski) (RETIRED) gentoo-dev 2002-01-24 13:13:38 UTC
is resume a problem with wget or within portage?
Comment 9 Geert Bevin 2002-01-29 01:09:01 UTC
Following are the portage fixes/additions that have been agreed upon for Gentoo
v1.0 :
1) USE toggles: USE="+gtk"; USE="-alsa"
2) auto-USE: each USE var can have a set of dependencies. We have an OPTION=""
var to turn on "auto-use if dep satisfied"(based on deps being satisfied)
3) external category definitions
4) SYNC variable
5) download reimplementation with resume
6) binary compatibility slot: install as /var/db/pkg/cat/${PN}-${SLOT} where
slot is by default ${PV} (ensures that for every version, only one rev will
exist by default)
Comment 10 Geert Bevin 2002-01-31 14:57:07 UTC
*** Bug 205 has been marked as a duplicate of this bug. ***
Comment 11 Geert Bevin 2002-02-04 16:42:00 UTC
binary compatibility slots have been implemented
Comment 12 Tod M. Neidt (RETIRED) gentoo-dev 2002-02-21 16:25:41 UTC
Hi!

A simple one.

The addition of a FFLAGS variable in /etc/make.conf and /etc/make.globals so
users can specify.

Comment 13 Daniel Robbins (RETIRED) gentoo-dev 2002-02-21 18:34:46 UTC
what if FFLAGS used for?
Comment 14 Tod M. Neidt (RETIRED) gentoo-dev 2002-02-22 17:30:22 UTC
FFLAGS the standard ENV variable for compiler options when using g77 (or any
other fortran compiler for that matter) They are analogous to CFLAGS and CXXFLAGS.

Would be useful for ebuilds that have fortran in the source.  Currently I have
either been hardcoding sane and or recommended defaults or just letting the
defaults in the Makefile stand.

Example ebuilds for packages having all or part of the source in fortran are
dev-lang/R
dev-lang/octave

In fact you might be suprised (I was) at which packages would use FFLAGS, I just
'grep -Hr FFLAGS /var/tmp/portage/*' and gnome-python and sylpheed pooped up
(wouldn't have figured :)

Comment 15 Daniel Robbins (RETIRED) gentoo-dev 2002-03-06 12:53:38 UTC
Tod, send me some good FFLAGS.  Should they be the same as CFLAGS?
Comment 16 Tod M. Neidt (RETIRED) gentoo-dev 2002-03-07 23:27:46 UTC
Hi!

I was afraid somebody would ask me that :)

I would hazard to suggest the following for a reasonable optimization default
assuming that x86 is the only cpu architecture Gentoo will support in the near
future. Most FORTRAN flags tell g77 how to compile for the FPU and depends on
the architecture and which IEEE floating point calculation standards it conforms to.

FFLAGS="-march=<arch> -O3 -pipe -malign-double -funroll-loops"

The first three are the standard flags currently used, i.e. CFLAGS and CXXFLAGS.
The -malign-double (Note: specific to x86 processors *only*) and -funroll-loops
(only "DO" loops) flags are especially applicable to FORTRAN programs.  These
are the only two flags mentioned that improve performance in general.  The other
flags only *might* improve performance depending on the code and cpu
architecture.  A complete listing of compiler flags useful for FORTRAN can be
found at

http://gcc.gnu.org/onlinedocs/gcc-2.95.3/g77_9.html#SEC18

or, alternatively, the analogous node in 'info g77'

I wouldn't characterize myself as an expert on FFLAGS.  If someone else has
comments or suggestions, I would be interested in hearing them.
Comment 17 Tod M. Neidt (RETIRED) gentoo-dev 2002-03-08 10:38:27 UTC
Concerning optimization flags, I did some poking around and found some
interesting information.

First an article on Linux Journal:
http://www.linuxjournal.com//article.php?sid=5886

which referenced "Securing and Optimizing Linux"
http://www.linuxdoc.org/LDP/solrhe/Securing-Optimizing-Linux-RH-Edition-v1.3/gen-optim.html

6.1. The /etc/profile file, has recommended flags, but the next few sections
after have interesting benchmark results.

But it turn out all this info is derived from the http://gcc.gnu.org/benchresults/

Pretty graphs are found at http://gcc.gnu.org/benchresults/cov.html which breaks
out benchmarks by C++, FORTRAN, etc.

It is interesting how the recommended optimization flags change subtly going
from i486 --> i586 --> i686
Comment 18 Spider (RETIRED) gentoo-dev 2002-03-19 20:48:31 UTC
perhaps a LDFLAGS would be nice? (See KDE-thread in gentoo-users about performance)

reasonable (?) would be  -z combreloc