#131 closed defect (duplicate)
Instead of running MAKEDEV, run the mknod commands
| Reported by: | Owned by: | ||
|---|---|---|---|
| Priority: | lowest | Milestone: | |
| Component: | Book | Version: | CVS |
| Severity: | normal | Keywords: | |
| Cc: |
Description
This has been suggested in an email to lfs-discuss.
From: David Anselmi <anselmi@…> To: lfs-discuss@… Subject: Re: LFS MAKEDEV script Date: Fri, 17 Aug 2001 22:07:56 -0600 First paragraph from the email:
To me, seems that the makedev script is not very LFS-like. It wouldn't surprise me to have a page in the book that says "now make all your device files with the following mknod commands" and lists the minimal set of device files (perhaps with a mention of the handy makedev script). After all, LFS is a minimal install that you build what you want on top. Why shouldn't it provide a minimal set of devices and you build what you need on top?
---
Would be great to have in the LFS-3.0 release. I'll give it a start immediately (as soon as I finish building here in chapter 5)
Attachments (2)
Change History (19)
comment:1 by , 23 years ago
| Owner: | changed from to |
|---|---|
| Status: | new → assigned |
comment:2 by , 23 years ago
comment:3 by , 23 years ago
Use this email to lfs-discuss with the explantions:
From: Sergey Ostrovsky <sostrovsky@…> To: lfs-discuss@… Subject: Interface userland - kernel device on the nutshell ( use to be LFS MAKEDEV SCRIPT ). Date: Sun, 19 Aug 2001 11:02:44 -0400
comment:4 by , 23 years ago
| Resolution: | → later |
|---|---|
| Status: | assigned → closed |
For after lfs-3.0 - no more time to mess with this
comment:5 by , 23 years ago
in the backup MAKEDEV script, include Matthias's patch that will try a regular chown first and if it fails (missing Glibc), parse the /etc/passwd file directly.
comment:6 by , 23 years ago
| Resolution: | later |
|---|---|
| Status: | closed → reopened |
by , 23 years ago
| Attachment: | MAKEDEV.diff added |
|---|
patch to makedev so it works before Glibc is installed
by , 23 years ago
| Attachment: | make_devices.bz2 added |
|---|
alternative to MAKEDEV to only install what you need, not the ~1027 you don't need
comment:7 by , 23 years ago
| Priority: | normal → high |
|---|
comment:8 by , 23 years ago
| Priority: | high → normal |
|---|
comment:9 by , 22 years ago
| Priority: | normal → highest |
|---|
Just put that patch in for the LFS-4.0 release so MAKEDEV can run before Glibc. We'll worry about alternate to MAKEDEV some other time.
comment:10 by , 22 years ago
MAKEDEV has been moved to before glibc and the manual creation/deletion of /dev/null removed. Shall I close this bug until after LFS-4.0 is released?
comment:12 by , 22 years ago
| Resolution: | → fixed |
|---|---|
| Status: | reopened → closed |
closing. may need re-opening after LFS-4.0
comment:13 by , 22 years ago
| Resolution: | fixed |
|---|---|
| Status: | closed → reopened |
comment:14 by , 22 years ago
| Priority: | highest → normal |
|---|
comment:15 by , 22 years ago
Just a note that the "Creating the password and group files" section mentions the groups created are those that the MAKEDEV script uses. Once the transition to mknod is made that page will need updating, possibly along with the particular group creation commands.
comment:16 by , 22 years ago
| Priority: | normal → lowest |
|---|
comment:17 by , 21 years ago
| Resolution: | → duplicate |
|---|---|
| Status: | reopened → closed |
* This bug has been marked as a duplicate of 78 *
Right. Sorry for being slow but I'm AFK for a few days at the moment. Just my comments on this: IMHO we should at least leave the MAKEDEV script as an option. Personally, although I want to build my system from scratch, creating all of the devices using mknod just seems pedantic. For example, those using systems which don't use devpts are going to have to create pty[p-za-e][0-9a-f] and tty[p-za-e][0-9a-f] manually which, even using for where possible loops is a lot of devices.. I seem (although I've only had time to scan my email quickly) to be going against the mailing list grain on this one but I just think that it's going to be a cause of a lot of support requests... As I say, maybe we should give both methods. Just my tuppence.