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

[James Blackwell]

(btw, in case anyone is curious as to my quoting style, slrn is
braindead when it comes to figuring out the recipient. Being as lazy as
anyone else, I just fix the attribution by hand)

> James Blackwell wrote:
>> Hi all.
>> 
>> The time quickly approaches for a 1.2.2rc1 release. If you have
>> patches that you want to see in 1.2.2 and they're not here, then talk
>> to me quickly.
>> 
>> Some quick notes about whats in here: 
>> 
>>   1. There is a test-update testcase that is currently failing. For
>>   some reason, the test case fails if address@hidden is not
>>   registered I have no idea how to fix it.
 
Aaron Bentley wrote: 
> Last time I checked, that test case was an untagged source file.  This
> should probably be fixed.

Already fixed. I was putting off the d/a until the case was fixed, but
it was annoying Tom enough that he fixed it. So I took his fix. ;) 

>
>>   2. The tag boundary detection code has illuminated some long term
>>   tla misbehavior that nobody seemed aware of. As such, panicing was
>>   too strong a step to take, and the panic has been reduced to a
>>   warning.
>
> That warning should not happen at all if the tag isn't really being
> truncated.  I believe that was Duffy's case?

I'm not sure. There's been a lot of confusion. In some cases, it sounds
like rather than trying bottom, then top, the top is being checked
first. (which is a bug in and of itself). Regardless of the case, the
panic is just making apparent previous silent behavior.

Earlier today, I changed the panic to a warn for jao, and he reported
that tla *segfaulted*. So I'm increasingly starting to think that these
patches are going to have to be pulled prior to rc1. We can revisit the
issue with a greater level of detail later.

> And maybe the warning belongs in tree-lint, rather than the main
> inventory.

Yeah, that's one possible answer. Another answer (credit: lifeless) is
to warn on inventory, and panic in tree-lint. Even this might be a bit
strong though.

I think that whatever we do, we have to somehow notify the users that
tla has been truncating their tags all along.  In some of the edge
cases, any modification that changes the offset of the tag changes the
tag id, which results in an a/d pair.




-- 
James Blackwell          Try something fun: For the next 24 hours, give
Smile more!              each person you meet a compliment!

GnuPG (ID 06357400) AAE4 8C76 58DA 5902 761D  247A 8A55 DA73 0635 7400