Re: [Gnu-arch-users] Upcoming release of 1.2.2rc1

[Robert Anderson]

--- Original Message ---
From: address@hidden (James Blackwell)
To: address@hidden
Subject: Re: [Gnu-arch-users] Upcoming release of 1.2.2rc1

>> This is crazy.  There is a bug in released version that prevents
>> people from _checking out their code_ from the archive, in a case
>> that will happen quite frequently on a code base of any
>> appreciable size.
>>
>> 1.2.1 needs an emergency patch and re-release, if you want people
>> to take it seriously as a piece of software ready for anything
>> but the most casual experimentation.
>
>Pardon? The 1.2.1 I released and signed at
>http://release.gnuarch.org/tla-1.2/1.2.1/ doesn't have this
"problem."

Well there's the first bug.

You are releasing "tla 1.2.1" from "gnuarch.org" and then
disavowing responsibility for anything in the "tla 1.2.1" release
from ftp.gnu.org.

Either you are releasing GNU tla 1.2.1 or you are not.  If you're
not, please don't make your releases from a site with "gnu" in
the title.  That's just misleading and confusing.

>Just so that you understand where we are, tla panicing is *not*
the bug.

I disagree.  I think it is a bug.

>The patch just makes apparent a a well hidden, but serious id
corruption
>bug thats been happening for a long while now, probably since
tla inception.

Maybe you can demonstrate this bug, then, because I have tried
tla inventory --ids using an older tla in the bug I reported and
there was _no truncation_, and this is the second time I've
reported those results.

>Regarding pushing the release cycle harder -- That it is a mistake.

I think waiting for a "release cycle" for a critical show-stopper
bug in a release is a mistake.  I think those should be fixed
before any other work gets done.

>So a lot of people are being constructive, trying to figure out
a safe,
>consistant way to make the bug go away with as little pain for
you as
>possible.

This is not about me.  This is about the quality of tla releases
and the release process.

 There's several possible solutions on the table, but they all
>have negative consequences. Even in the best case scenarios, we're
>going to have to subject you to an add/delete pair for every
file with a
>corrupted id.
>
>This isn't because the ids were corrupted in 1.2.1. They've been
>corrupted all along.

Please demonstrate this, because only one of two things is
possible here:

1) there's a bug in tla inventory --ids which is masking this
purported truncation.

2) there's all kinds of version skew between whatever releases
you are working with and the official releases (that I work with)
that are going to cause no end of confusion if there isn't a
serious effort to differentiate whatever you're releasing from
"GNU tla".

Bob