Jonathan Nieder wrote:
> Adam D. Barratt wrote:
>> * the preinst isn't idempotent
>> - from a few quick tests, it looks like this could be resolved by changing
>> the execution order to "cp, ln, divert" rather than "divert, cp, ln"; I
>> may have missed something, however
>
> Technically one would
Hi,
Adam D. Barratt wrote:
> This bug report seems to have branched in to a couple of issues relating
> to dash's use of diversions and its use as /bin/sh, rather than the
> original issue of whether the preinst should be moved to something that
> doesn't rely on /bin/sh working; some of these sh
Hi,
This bug report seems to have branched in to a couple of issues relating
to dash's use of diversions and its use as /bin/sh, rather than the
original issue of whether the preinst should be moved to something that
doesn't rely on /bin/sh working; some of these should probably be moved
elsewhere
Adam D. Barratt wrote:
> The key thing is that if the move to dash-as-default - or an upgrade of
> dash whilst it is the default shell - fails, that should not leave the
> system without a working /bin/sh. Having it not be a shell script seems
> a reasonable why of achieving that. :-)
Yes, makes
On Fri, 2010-10-08 at 23:02 -0500, Jonathan Nieder wrote:
> Presumably the idea is that before bash is updated, its /bin/sh is
> always diverted, so the removal of bash's copy of /bin/sh causes
> no harm.
>
> I don't see why dash's preinst urgently has to be rewritten in C
> (though I don't disagr
On Fri, Oct 08, 2010 at 11:02:22PM -0500, Jonathan Nieder wrote:
> A somewhat orthogonal task
> ~~
> IV. Introduce a new package to take responsibility for the /bin/sh symlink.
>
> Preinst (not necessarily in this order):
> 1. Removes all diversions
> 2. Set
Jonathan Nieder dixit:
>Thoughts? Questions? Advice?
Other shells would need to be changed to cope with that as well.
The situation is a bit tricky at the moment, because mksh also
contains code to manage a diversion, which only works if dash
doesn’t, though.
Worse, on upgrade from lenny to sq
Gerrit Pape wrote:
> You know, I know how to do it.
Okay, I did some digging in the archives. Sorry to those who
already went over this all.
debian-package-dash-getv...@list.smarden.org
debian-package-dash-getv...@list.smarden.org
debian-package-dash-getv...@list.smarden.org
In April, Ge
Gerrit Pape wrote:
> I say, just do it,
> finally. The dash package will benefit from this too, if you manage to
> do it right.
Well said, that.
> You know, I know how to do it. BUT, in my case, I'm afraid the Debian
> people won't like it, and, wh
On Thu, Oct 7, 2010 at 20:11:21 +, Thorsten Glaser wrote:
> Support mksh as /bin/sh and RoQA dash NOW!
> ‣ src:bash (235 (253) bugs: 0 RC, 177 (192) I&N, 58 (61) M&W, 0 F&P)
> ‣ src:dash (72 (82) bugs: 4 (9) RC, 49 (54) I&N, 18 M&W, 1 F&P)
> ‣ src:mksh (2 bugs: 0 RC, 0 I&N, 2 M&W, 0 F&P)
>
I
On Thu, Oct 07, 2010 at 08:11:21PM +, Thorsten Glaser wrote:
> See my signature. Anyway, I think we should have some common, shared
> code across all shells that are eligible as /bin/sh in Debian, and
> maybe manage /bin/sh as non-diverted non-packaged file, means symbolic
> link. The shells’ m
Jonathan Nieder dixit:
>[1] except that preventing sysadmins from making /bin/sh point where
>they want is pretty unfriendly.
ACK.
>Thoughts?
See my signature. Anyway, I think we should have some common, shared
code across all shells that are eligible as /bin/sh in Debian, and
maybe manage /bin
(+cc: shell people)
Jonathan Nieder wrote:
> 3. If dash diverts /bin/sh, then bash's /bin/sh will be diverted
> at unpack time: no file conflict. If bash diverts /bin/sh, then
> dash's /bin/sh will be diverted at unpack time: still no file
> conflict. But what if the sysadmin diverts /bin/sh wi
Hi Raphael!
Some quick questions about dash.preinst:
1. The diversions for /bin/sh and /usr/share/man/sh.1.gz are handled
separately. Is that intentional? What is supposed to happen if
someone diverts /bin/sh but not the manpage (for example because a
maintainer script was interrupted)?
2. The
On Saturday 17 July 2010 12:48:09 Jonathan Nieder wrote:
> Will there be a vcs repository for interested people to peek at
> your changes?
Yep, it is already available at
http://git.debian.org/?p=users/geissert/dash.git;a=summary
The changes I've made so far (for the move) are not in that reposit
Raphael Geissert wrote:
> The preinst script needs more changes for
> a successful move of /bin/sh from bash and dash to only dash.
> That's already on my debcamp/debconf schedule.
Thanks, Raphael. That’s good to hear.
To clarify: I was only suggesting that any change as big as rewriting
the d
Hi,
On Saturday 17 July 2010 08:24:52 Marek Kubica wrote:
> In the course of todays BSP in Munich, I created an NMU patch, you'll
> find it attached. I applied David's patch unchanged. I hope we can
> upload this at the end of the BSP.
Please hold those changes (or if you insist, upload them to e
On Sat, Jul 17, 2010 at 15:24:52 +0200, Marek Kubica wrote:
> tag +pending
> thanks
>
> Hi,
>
> In the course of todays BSP in Munich, I created an NMU patch, you'll
> find it attached. I applied David's patch unchanged. I hope we can
> upload this at the end of the BSP.
>
> regards,
> Marek
>
Hi Marek,
Marek Kubica wrote:
> In the course of todays BSP in Munich, I created an NMU patch, you'll
> find it attached.
I would prefer to see this applied to experimental. You can remove
debian/diff/0006--INPUT-exit-127-if-command_file-is-given-but-doesn-t.diff
to fix Bug#586807 while at it.
tag +pending
thanks
Hi,
In the course of todays BSP in Munich, I created an NMU patch, you'll
find it attached. I applied David's patch unchanged. I hope we can
upload this at the end of the BSP.
regards,
Marek
reverted:
--- dash-0.5.5.1/debian/dash.preinst
+++ dash-0.5.5.1.orig/debian/dash.prei
retitle 546528 dash's preinst shouldn't rely on /bin/sh
tag 546528 +patch
thanks
* Matthias Klose [2009-10-22 20:08]:
> changing that to a pre-dependency, but dash's preinst shouldn't rely
> on /bin/sh. bash's preinst is a C binary for that reason.
Here's a patch that makes dash's preinst a C bi
21 matches
Mail list logo
