[OmniOS-discuss] Bug in pkg/server on OmniOS when the publisher name contains a dash

[Volker A. Brandt]

Hi Dan!


> > This is paraphrased from a mail that I had sent to Dan privately
> > on March 31st.  Unfortunately, the problem still persists in
> > 151014.
> 
> Dug through my mail, and I didn't see this note.  I may have lost
> it, but gmail is usually good about not letting you get rid of
> mails.  :(

Well, I sent it to @omniti.com.  My guess is you don't see any of
my mails because they are filtered away.  But no matter, as long as
the discuss list works, that's fine.
 
> I'm seeing a different variant of this error using a toy repo I set
> up.  It has to be an http repo using pkg.depotd.  If I use a
> file:/// URL, it appears to work.

Exactly.  Good to know that I am not imagining things.
 
> The OmniOS version of pkg(5) is a downstream of OpenIndiana's, which
> is in turn a downstream of Oracle's, but last synched in 2013
> (because Oracle uses hg, and git from hg is annoying).

Interesting, wasn't aware of that.  So is it really just a matter
of mercurial->git conversion, or has the OI/Omni version of pkg
diverged too much?

if the former, maybe one of the many tools could be leveraged that
pull stuff out of hg repos and push it into git?

> The github
> URL is https://github.com/omniti-labs/pkg5/ if you wanna poke around
> in the source we use.  NOTE we have per-release branches starting
> with r151014, but the "default master" is called "omnios" because
> we're a downstream child of another downstream child.

Thanks for the pointer, but I guess it would take me ages to spot,
let alone fix the bug.
 
> I don't know of anyone who uses dashes in their publisher name.

Well, now you know. :-)  BTW IHAC who uses lots of dashes in lots
of publisher names under Solaris... just a matter of taste I guess.

> wonder if this bug was around a while, and that's why it's
> "ms.omniti.com" instead of "omniti-ms", for example?

Sounds likely.  I wouldn't know. :-)
 
> ANYWAY, my toy server with an empty repo had this in its logs:
> 
> 127.0.0.1 - - [20/Apr/2015:13:32:46] "GET /versions/0/ HTTP/1.1" 200 179 ""
> "pkg/1427212657 (sunos i86pc; 5.11 omnios-fbd6dc7; full;
> pkg)"

Interesting.  Where does that log come from?  Do you have an Apache
with redirect rules sitting in front of your repo?  I did play around
with redirects and rewrite rules, but wasn't really happy, so I decided
to concentrate on one problem at a time.  My Apache logs show similar
entries when I tell the IPS client to use port 80 (where an Apache is
listening on the pkg server host):

192.168.xxx.xxx - - [20/Apr/2015:17:58:10 +0200] "GET /my-repo/versions/0/
HTTP/1.1" 404 220 "-" "pkg/1427212657 (sunos i86pc; 5.11 omnios-170cea2; full;
pkg)"

...but the pkg/server:my-pub instance listens on another port, hence
the 404 codes.


Going forward, what is the likelihood of OmniOS adopting a more recent
version of the Oracle IPS gate?  I don't really think I can offer much
help (ENOTIME, ENOCLUE, etc.) but IMHO it would be well worth while if
it fixes that bug.


Regards -- Volker
-- 
------------------------------------------------------------------------
Volker A. Brandt               Consulting and Support for Oracle Solaris
Brandt & Brandt Computer GmbH                   WWW: http://www.bb-c.de/
Am Wiesenpfad 6, 53340 Meckenheim, GERMANY            Email: vab at bb-c.de
Handelsregister: Amtsgericht Bonn, HRB 10513              Schuhgröße: 46
Geschäftsführer: Rainer J.H. Brandt und Volker A. Brandt

"When logic and proportion have fallen sloppy dead"