1 |
On Sunday 18 April 2010 19:37:40 Tanstaafl wrote: |
2 |
> On 2010-04-18 1:05 PM, Alan McKinnon wrote: |
3 |
> > On Sunday 18 April 2010 17:38:19 Tanstaafl wrote: |
4 |
> >> I must be missing something obvious... |
5 |
> >> |
6 |
> >> When updating gcc, and thus emerging -e both system and world |
7 |
> > |
8 |
> > Why are you doing that? |
9 |
> |
10 |
> Because it wasn't a minor bugfix update? I'm going from 4.1.2 to 4.3.4. |
11 |
|
12 |
So what? |
13 |
|
14 |
Is there an ABI change between 4.1.2 to 4.3.4 that causes apps built with |
15 |
4.1.2 to break on 4.3.4? |
16 |
|
17 |
> > > Have you verified that you have in fact updated gcc from one version |
18 |
> > to another with an incompatible ABI? |
19 |
> |
20 |
> Since the GCC upgrade guide doesn't mention a way to do this, no, I |
21 |
> haven't... so how does one verify this, Obi-Wan? ;) |
22 |
|
23 |
Understand what the gcc upgrade guide is: |
24 |
|
25 |
a quick simple guide for user who don't know tool chains inside out, that give |
26 |
the minimal sequence of commands that is GUARANTEED to NOT leave the user in |
27 |
the lurch. It's not the minimum possible sequence of commands to do the |
28 |
upgrade, because most users don't know how to tell the difference. Here's the |
29 |
logic: |
30 |
|
31 |
"If we imply that you should merge -e world with a gcc upgrade, my inbox will |
32 |
not fill with bugs from users who don't know the inner guts of the toolchain". |
33 |
|
34 |
The Guide is not the best possible guide that can be. It is the guide that |
35 |
causes the least support questions from users. |
36 |
|
37 |
> > There is a false idea floating about that gcc updates require |
38 |
> > updating -e world. |
39 |
> > |
40 |
> > This is not true. |
41 |
> |
42 |
> I've heard this before, but could never get a definitive answer on how |
43 |
> to know for sure when you do and don't need to... |
44 |
|
45 |
When the gcc devs release a new version, they will announce that there is an |
46 |
ABI break, and what it applies to. It doesn't happen often, but when it does, |
47 |
the gentoo maintainer ought to put a note in the ebuild. |
48 |
|
49 |
They are quite diligent about this, much like the KDE/Qt devs do the same |
50 |
thing when Nokia breaks Qt compatibility. |
51 |
|
52 |
> >> where are the info/warn/errors logged? I see a whole bunch of |
53 |
> >> separate/individual logs ion /var/log/portage/elog, but what I'm |
54 |
> >> looking for is a single, easily readable log of all of the result |
55 |
> >> info/warn/errors for each package. |
56 |
> > |
57 |
> > Configure it in /etc/make.conf, mine looks like so: |
58 |
> > |
59 |
> > PORT_LOGDIR=/var/log/portage |
60 |
> > PORTAGE_ELOG_CLASSES="warn error log qa" |
61 |
> > PORTAGE_ELOG_SYSTEM="save" |
62 |
> |
63 |
> I have the last two, but didn't have the first - just added it... |
64 |
|
65 |
I think that's the default. I have it just to be explicit. |
66 |
|
67 |
> |
68 |
> So, does adding the first entry provide an overall emerge.log with just |
69 |
> all of the messages in it that I can review? |
70 |
> |
71 |
> > Adapt to you own needs. |
72 |
> > |
73 |
> > If you don't have this configured, then the logs are not stored. To |
74 |
> > see them again, read the ebuilds or re-run emrge |
75 |
> |
76 |
> Like I said, I have a bunch of *individual* logs (for individual |
77 |
> ebuilds)... I was hoping for something a little easier to manage/read, |
78 |
> all in one file... |
79 |
|
80 |
for I in $(ls -1rt /path/to/emerge/logs/*) ; do cat ${I} >> some_log_file ; |
81 |
done |
82 |
|
83 |
-- |
84 |
alan dot mckinnon at gmail dot com |