[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: com_err

>>>>> "Chris" == Chris Chiappa <griffon+heimdal-discuss@snurgle.org> writes:

    Chris> As anyone who's played with Kerberos and friends knows, I'm
    Chris> sure, there are way too many software packages out there
    Chris> that bundle com_err and/or ss with them.  Generally this is
    Chris> more an annoyance than anything else; you can usually trick
    Chris> programs into using another's so you don't end up with too
    Chris> much clutter on your hard drive.  I'm having a bit of
    Chris> challenge achieving this with Heimdal however.
    Chris> Specifically, I'm trying to build some local packages for a
    Chris> Debian system.  Debian (and others, Redhat for one)
    Chris> includes a "canonical" libcom_err derived from the
    Chris> e2fsutils.  Unfortunately,

    Chris> (1) Heimdal cannot use the Debian compile_et because it
    Chris> uses a "PREFIX" extension in .et files which isn't
    Chris> supported by the MIT-derived compile_et.

    Chris> (2) Other programs which wish to use Heimdal's generated
    Chris> _err.h files won't be able to because they require
    Chris> com_right.h, requiring Heimdal's com_err to be installed.

    Chris> This leaves one in the bad position of being forced into
    Chris> installing two incompatible versions of com_err on their
    Chris> system.

    Chris> I'm looking for suggestions on how one would resolve this.
    Chris> The options that spring to mind are

    Chris> (1) Convince Debian to upgrade to the (presumably) superior
    Chris> com_err from Heimdal as the standard.  I'm assuming that
    Chris> the Heimdal stuff would be sufficiently capable of
    Chris> supporting all the programs which rely on MIT-like com_err
    Chris> behavior. While this would probably get Debian a better
    Chris> com_err it seems to add a lot of overhead for such a
    Chris> trivial package.  Is the KTH com_err available as a
    Chris> separate package from anywhere?

    Chris> (2) Modify the Heimdal .et files so they don't rely on the
    Chris> KTH extensions.  While the PREFIX extension seems nice and
    Chris> sensible, removing it would increase compatibilty.

    Chris> (3) As an alternative to (2) perhaps an autoconf switch for
    Chris> Heimdal which could (At the user's request) use the
    Chris> installed com_err and preprocess the .et files to get rid
    Chris> of PREFIX.

    Chris> (4) Modify the KTH compile_et so that it generated
    Chris> MIT-compatible _err.h files (ie that only #include
    Chris> <com_err.h>).  Seems ugly as it looks offhand like the data
    Chris> structures they want are different.

    Chris> I'm somewhat keen on (2) or (3) myself; (2) could be kept
    Chris> as a local modification without too much pain.  It might be
    Chris> nice to spare others these problems however.  Comments?
    Chris> Criticisms?  Suggestions?  Flames?  I'm all ears.

I bought up this (and some other related issues) with some Debian
people, including Sam Hartman, the Debian maintainer of MIT
Kerberos. I have attached his response.

For com_err, please get Heimdal people in touch with Ken Raeburn
<raeburn@mit.edu>; he is working on com_err unification.  He is aware
of the Heimdal situation, and may have talked to Assar when Assar was
last in Boston.  Still, Both Assar and Ken are generally busy and
smashing the two of them together would probably help.

2) Assar agreed (verbally) that Heimdal would not install a libsl or
   libcom_err or libdes if it was compiled against an installed
   kerberos4kth.  I don't know if he ever implemented that.

3) No idea what to do about libsl; no one has investigated libss (MIT)
   vs libsl conflicts.

Brian May <bam@snoopy.apana.org.au>