Summary/Minutes from today's FESCo Meeting (2012-06-18)

[Kevin Kofler]

Kevin Fenzi wrote:
> 18:23:37 <nirik> #topic ticket 868 F18 Feature: MiniDebugInfo -
> https://fedoraproject.org/wiki/Features/MiniDebugInfo
> 18:23:37 <nirik> .fesco 868
> 18:23:42 <zodbot> nirik: #868 (F18 Feature: MiniDebugInfo -
> https://fedoraproject.org/wiki/Features/MiniDebugInfo) – FESCo -
> https://fedorahosted.org/fesco/ticket/868
> 18:24:03 <nirik> mmaslano was 0 in ticket.
> 18:24:24 <nirik> I'm -1.

First of all, thank you Kevin for having been the only one stepping up 
against this nonsense!

> 18:24:28 <limburgher> Seemed. . .mildly contentious. . .
> 18:24:34 <mitr> I think the only technical objection was that spins won't
> fit on a CD,

The "only" technical objection which is a blocker as far as we're concerned. 
Image sizes are even a release blocker! Nobody decided that spins should NOT 
be CD-sized.

There was also the additional objection that the feature brings no practical 
gain, because -debuginfo is already available and the Retrace Server is 
available. Later in the discussion, you also pointed out that ABRT probably 
won't even use this information! So why ship it? Bloat just for the sake of 
bloat?

> but when repeatedly asked nobody came up with an example user/ambassador
> saying that CDs are needed

Huh? Examples have been given in the mailing list threads. There's also the 
constraint that all the spins in both 32-bit and 64-bit editions need to fit 
on one DVD (the Multi DVD), which very much does matter to the ambassadors.

> 18:24:54 <mezcalero> nirik: out of curiosity, why do you say -1?
> 18:25:17 <notting> i am +1. obviously would need to be coordinated with
> any dmz-related rebuilding
> 18:25:21 * pjones is also +1
> 18:25:33 <nirik> I don't think the extra size is worth the gain. I think
> we can/should work on making abrt better.

Right.

> 18:25:34 <hughsie> i don't know if i should comment, but from my role as
> an upstream maintainer, this would be really useful for real life
> debugging from end users...

That's what -debuginfo packages are for!

> 18:25:44 <limburgher> mitr: Would dwarf compression help that?
> 18:25:48 <jwb> i'm marginally +1
> 18:25:55 <pjones> limburgher: no - different set of fields

Even if it were affecting the mini-debuginfo fields (it doesn't), 
compression wouldn't help at all because the live images are already 
compressed. Even special-case compression has at best a very small effect 
after xz is run on it.

> 18:25:56 <mitr> limburgher: jakub said above that it is orthogonal
> 18:25:57 <t8m> as long as it does not replace full backtraces reported in
> ABRT I'm ok with it so marginally +1
> 18:26:11 <mitr> AFAICS this "only" benefits developers and reading crashes
> in unexpected components, but given Fedora's target arudence, +21
> 18:26:14 <mitr> +1 that is
> 18:26:16 <limburgher> mitr, pjones:  Missed that, thanks.
> 18:26:22 <limburgher> Ok.  +1
> 18:26:24 <mjg59> I lean +1
> 18:26:43 <mitr> nirik: My best guess is that abrt will just ignore this.

So just useless bloat.

> 18:26:57 <nirik> #agreed Feature is approved. (+7/-1/0)
> 18:27:01 <_jakub_> I don't see what minidebuginfo gives us as advantage
> for bugreporting, in theory all you need is addresses (relative to start
> of DSOs) and their build-ids, everything can be recomputed from that

So there, THE expert on the domain tells you the bloat is totally useless, 
but it's too late, you already rushed out a vote of a bunch of "marginal" 
+1s. :-/

> 18:27:03 <limburgher> I wish spins could stay on CDs, but then I still
> want CD and floppy install media.  <stares wistfully into the past>

Floppies are clearly too small. CDs are not. They'd fit all we need if it 
weren't for the completely useless bloat added by "features" such as this.


I also object to the fact that the feature owners have spun the feature 
proposal in a deliberately misleading way:
* "increase quality of bug reports"? – ABRT won't even use this!
* "allow easier support for profiling and user space tracing"? – Why 
wouldn't one just install -debuginfo for those tasks? It's not what a user 
needs to do every day.
* misleading percentages – see Talk page
* Also note that for shared objects, we already HAVE symbols for the 
exported functions (for free!), which is often enough!

        Kevin Kofler