Re: new document
Matej Kosik wrote:
The document was written by total newbie which dug the archives.
I wanted to compile gnumach/hurd from source mainly because I wanted
to see that it really works.
Your document really needs changes to be useful for others.
Ognyan Kulev wrote:
> - It would be better if you mention that using Debian mirrors is
preferred.
sorry I don't understand what you are saying, please be more concrete
You use http.us.debian.org in some URLs, but it is much better if users
replace it with their nearest mirror listed in
http://www.debian.org/misc/README.mirrors
I wasn't aware of `gnumach-1.3/i386/README-Drivers' until now.
I will definitely mention it in the document, who will need
more information, will know where to find it.
> - Compiling GNU Mach requires the `mig' and `patch' packages too.
I assumed running `Debian GNU/Hurd' (as stated in introduction)
`mig', `patch' and other development packages are not included in the
tarball of Marcus so they must be installed manually. This can be
non-trivial for someone that is not familiar with Debian. I use
`apt-get install gnumach-dev hurd-dev libc0.3-dev mig gcc make autoconf
automake gdb cvs patch sharutils' for that.
> - Using serverboot is deprecated, although it still works.
what is severboot?
This program starts Hurd's essential servers like `proc' and `exec' but
it's deprecated. The new way is using `ext2fs.static' (and `ld.so') as
described in Neal's installation manual.
(where do I mention it?)
In `Installing GNU Mach 1.x' section. And why in `Installing GNU Mach
2.x' you have `kernel oskit-mach' and `module gnumach' in menu.lst. I
never used oskit mach but AFAIK there is no need for `module gnumach'.
> - Why do you need -D 'Sun Jul 21 00:45:02 UTC 2002' when getting GNU
> Mach 2.x, or I slept some news about GNU Mach :) ?
gnumach-2.x wasn't officially released (correct me if I am wrong) yet
so I simply refer to SOME cvs state I personally tried rather then
letting user always checkout the CURRENT version (whose properties
I can't predict).
>
> Do you have some better idea which version of gnumach-2.x to
> refer from CVS?
The default branch is GNU Mach 2.x and all fixes to GNU Mach will go
there (and probably into 1.x branch). It is definately better to not
use `-D'.
> - It would be good to mention the `make kernel-ide+ethernet_ne2000'
> ability in GNU Mach 2.x.
(the `kernel-ide+ethernet_ne2000' (and the like) is a mistery
which I don't understand)
AFAIK it is not documented anywhere. (I learned about it from a mail by
Jeff Bailey who could happily miss this detail in his mail. :) It is
mentioned in gnumach/Makefile.in:320. Instead of using `make kernel' to
build kernel, in OSKit Mach you have to use `make kernel-DRIVERS', where
DRIVERS is `DRIVER+DRIVER+...+DRIVER' (a list of drivers separated by
`+'). DRIVER can be one of:
- `ide'
- `floppy'
- `ethernet_ETHDRV' where ETHDRV is taken from
`oskit/oskit/dev/linux_ethernet.h'.
- `scsi_SCSIDRV' where SCSIDRV is taken from `oskit/oskit/dev/linux_scsi.h'.
and resulting the kernel was usable. The rest was not interesting to me.
I don't have ethernet card/net (or to what `ethernet_ne2000' is related)
Ethernet and SCSI drivers are important. For example, my computer has
only SCSI disks, no IDE. And I use Ethernet to download more Debian
GNU/Hurd packages.
?? Can one determine the driver one wants by `configure' option
as well as with `make' command line options?
OSKit Mach is not configured by `configure', as you saw above.
> - It is much better to use CVS Hurd and CVS GNU Mach. Of course, using
> CVS source requires gnu-20020816.tar.gz or later.
For various reasons I didn't upgraded the system from
Debian GNU/Linux H4 "Sid" - Unofficial hurd-i386 Binary-{1,2,3,4}
(20020517) to some new (preferable also CD) distribution
Referring to fixed snapshot of Hurd (and GNU Mach) is bad. And after
all, why you would want to compile Hurd yourself if not for for applying
its latest fixes? Using fixed locations for GNU Mach 1.x and OSKit is
somewhat acceptable but it is not for GNU Mach 2.x and the Hurd.
Regards
--
Ognyan Kulev <ogi@fmi.uni-sofia.bg>, "\"Programmer\""
Reply to: