6 May
2008
6 May
'08
4:38 a.m.
I think it would be a good idea to import the V1 / and /usr filesystem from
the s2 tape into the subversion repository, so that we can write a tools/
script to build "known good" rf0.dsk and rk1.dsk disk images.
In my private tree I have 1root/ and 1usr/, but perhaps we should have
something like fs/root/ and fs/usr. They need to be split so that my
mkfs tool can build separate disk images.
Comments?
Warren
6 May
6 May
4:48 a.m.
I think it would be a good idea to import the V1 / and
/usr filesystem from
the s2 tape into the subversion repository, so that we can write a tools/
script to build "known good" rf0.dsk and rk1.dsk disk images.
The rf0 disk is pretty manageable. The rk0 disk is 2.5M and might
be a little big for putting into the SVN (download 2.5M any time
someone makes any changes to it)... If its something that isn't
going to change, perhaps just putting "known good" disk images up
on the website would be better. The google site lets us have a download
section, and then there's always TUHS...
I think the bigger question here is what do we want the standard disk
image to look like? Are we building rf0 with s2/bin and s2/etc only?
Should it have the jun72 init on it? the jun72 sh? A "mount" utility?
Is rk0 a clean copy of s2/usr? What about the files that mkfs can't
currently put on the system due to size? What about file permissions?
Are we going to restore the old s2 permissions? Will there be any
new versions or hacks applied?
Do we want to put up prebuilt kernels? If so, which patches, if any,
should be applied? If we put up a cold kernel I imagine we also want
to put up a tape to build rf0, and the same questions apply.
I imagine we want to have a script that autogenerates each of the
images that we intend to put up.
Warren
Tim Newsham
http://www.thenewsh.com/~newsham/
4:57 a.m.
On Mon, May 05, 2008 at 04:48:46PM -1000, Tim Newsham wrote:
Sorry Tim. I meant putting the files which will go onto the images, not
the images themselves. Plus a tools/ script to build the images.
I think it
would be a good idea to import the V1 / and /usr filesystem from
the s2 tape into the subversion repository, so that we can write a tools/
script to build "known good" rf0.dsk and rk1.dsk disk images.
The rf0 disk is pretty manageable. The rk0 disk is 2.5M and might
be a little big for putting into the SVN I think the bigger question here is what do we want
the standard disk
image to look like? Are we building rf0 with s2/bin and s2/etc only?
Should it have the jun72 init on it? the jun72 sh? A "mount" utility?
Will there be any new versions or hacks applied?
Perhaps we do the same as we have done with patches:
fs/root == original files as per s2 tape
fs/usr == original files as per s2 tape
fs/newfiles == new files to add to images
fs/buildroot == modified root (not in svn)
fs/buildusr == modified usr (not in svn)
tools/xxx == script to build root and usr
What about the files that mkfs can't currently
put on the system due to size?
Current mkfs can write all the s2 files, including the large ones.
What about file permissions? Are we going to restore
the old s2 permissions?
We can modify mkfs to restore uid, time and V1 permissions.
Do we want to put up prebuilt kernels?
No, just document the method to build the kernels.
I imagine we want to have a script that autogenerates
each of the
images that we intend to put up.
I think that's the way to go: a script to build a kernel + a set of
suitable tape/disk image files for it. Make it a top-level script
that calls next-level scripts, so that we can still run tools/assemv7
if we want to build a kernel and not build a filesystem.
Warren
8:53 a.m.
On Tue, May 06, 2008 at 12:57:44PM +1000, Warren Toomey wrote:
Perhaps we do the same as we have done with patches:
fs/root == original files as per s2 tape
fs/usr == original files as per s2 tape
fs/newfiles == new files to add to images
fs/buildroot == modified root (not in svn)
fs/buildusr == modified usr (not in svn)
tools/xxx == script to build root and usr
I've imported fs/root and fs/usr into the svn repository with the files from
the s2 tape, and fs/new has Tim's etc/init so far. I haven't written any
scripts to make filesystem images yet.
Rather than use the file owner/permissions on the fs/ files, I'll modify
mkfs.c to read a table a la the rawperms table in tools/mktape.py, so that
the exact owner/perms get put into the V1 filesystem images.
Cheers,
Warren
11:35 a.m.
New subject: [Unix-jun72] mkfs.c modified to set correct file owner/perms
On Tue, May 06, 2008 at 04:53:29PM +1000, Warren Toomey wrote:
I've imported fs/root and fs/usr into the svn
repository with the files from
the s2 tape, and fs/new has Tim's etc/init so far. I haven't written any
scripts to make filesystem images yet.
fs/Readme now contains the correct file permissions for the files on the
V1 filesystem images. mkfs.c now reads this file and uses it as required.
tools/imgbuild builds rf0.dsk and rk0.dsk, using the fs/ files. I decided
to turn off the "world write" flag on all the files: I can't see why any
files should be world writable.
Unfortunately, I had to use rsync in tools/imgbuild so that I could
stop the .svn/ directories being copied into the images. Let me know if
that is OK, or if we can do it without rsync.
Cheers,
Warren
6042
days inactive
6042
days old
5 comments
3 participants
participants (3)
-
Brad Parker -
Tim Newsham -
Warren Toomey