Re: [gentoo-user] Upgrading to gcc 4.1: emerge -e world required?

[Alexander Skwar]

Hemmann, Volker Armin wrote:
> On Saturday 27 May 2006 19:58, Alexander Skwar wrote:
>> Richard Fish wrote:
>> > On 5/27/06, Alexander Skwar <listen@×××××××××××××××.name> wrote:
>> >> gcc update "Nothing needs to be done".... Sure... How much did the
>> >> person who wrote this check? "Hello World!" worked, and that's it?
>> >>
>> >> Sometimes this complete lack of QA is really pissing me off :(
>> >
>> > Stop using ~arch packages, or stop whining.
>>
>> No, I won't do neither. The GWN and the upgrade doc used to say,
>> that an upgrade is (basically) riskless.
>
> no it does not.

Yes, it does.

> You are talking bullshit.

Am I?

> GWN:
>
> The number of applications that do not compile with gcc-4.1 is extremely small
> now, and most users should not experience any problems with ~arch packages
> not compiling.

Yes, read it. I'm not complaining about packages which do not compile because
of gcc 4.1.1. I complained because KDE stopped working. I complained, because
I had to re-compile glibc, so that I could compile glib.

So, please YOU read again.

> Read it, understand it. It is hard, I know. But it does not say 'riskless'.
> Not even 'basically riskless'. Read again.

Yes, please do so - please read again. Make note of the sentence right
before the sentence, which you've quoted:

| The upgrade should be incredibly easy and require no additional work to
| install and use.

If that's not "basically riskless", then I don't know what basically riskless is.

> And the uzpgrade guide says:

The upgrade doc used to say, that upgrading from 3.4.x to 4.1.1 will be
painless (I don't know the exact words anymore, as the box has been
thankfully removed, which is VERY good).

> Generally speaking, upgrades to bug fix releases, like from 3.3.5 to 3.3.6,
> should be quite safe -- just emerge new version, switch your system to use it
> and rebuild the only affected package, libtool. However, some GCC upgrades
> break binary compatibility;

And it said, that 4.1.1 was supposed to be binary compatible to 3.4.6.

> in such cases a rebuild of the affected packages
> (or even whole toolchain and system) might be required.

And thus, a rebuild of world/tc/system wouldn't be required.

>> >  ~arch works most of the
>> > time, but it is a _testing_ branch.  Do you expect the devs to login
>> > to each and every Gentoo user's system to test a new package and
>> > ensure complete functionality before adding it to ~arch?
>>
>> Bullshit.
>>
>> I'd expect them to do testing and not give so bold statements
>> as "The upgrade should be incredibly easy and require no additional
>> work to install and use. " without making VERY sure, that this
>> is actually true.
>>
>
> and ~arch is the testing ground. Basic testing 'it works or it works not' are
> the hard-masked packages.

I guess we'll disagree about the level of "basic testing". IMO there should
be different levels for ordinary packages (say net-mail/safecat) and rather
low level system packages like gcc, glibc - and maybe everything in the sys-*
categories.

> Maybe you should calm down?

I am calm. Just don't tell me, that everything's fine, when it actually
isn't.

>> Did you yet re-compile Qt 3 and Qt 4? No?
>>
>> Then you're experiences just don't count. KDE broke on my
>> system, when I recompiled Qt. Before the recompile, KDE was fine.
>> As I've wrote in lengths on the bug report. Seems you've not read
>> it - why not? Why am I writing reports and *also* post links
>> here?
>
> oh, that is sooo surprising.

Yes, it is, isn't it?

> Most of the times, a qt-update requires
> recompiling kdelibs, base and network (and kdepim).

Where was there a Qt update?

So, thanks for agreeing with me

> Something that happens
> even without gcc-updates.

No, it doesn't.

>> Did you try to compile glib? No? Then I guess you've done no testing.
>
> if he does not have glib?

Then he installs it.

>> Or what kind of testing have you done?
>
> enough for his system?

Not enough to experience the errors which have been reported here.

>> >  Since these are all heavy
>> > C++ users, I am sure that for my (pure ~x86) system, there are no
>> > issues.
>>
>> Congrats. It's not only me who's having problems.
>>
>
> no, you are not the only one, but you are one who makes a lot of fuss about
> problems, that are easy to solve

Pardon, but an "emerge -e world" is not easy to solve. Sure, it's not
that hard, but it takes so very long and because of that, it's to be
avoided like the plague.

And also pardon me, when I'm annoyed because of too bold statements
which turn out to be wrong. If it says "no problems expected", then
that's what I expect. I don't expect to run into deep problems. And
the GWN and upgrade doc clearly stated, that there were no problems
to be expected.

Heck - the GWN still says so. You should just read it yourself. Or
let me read it to you:

| The upgrade should be incredibly easy and require no additional
| work to install and use. The number of applications that do not
| compile with gcc-4.1 is extremely small now, and most users should
| not experience any problems with ~arch packages not compiling.

So, "emerge -e world" is "no additional work" for you? Well, guess
we've got differing oppinions in this case then.

> and even happen without any gcc updates  -

Those problems? Seldom.

> and you should have learnt to deal with a long time ago.

I deal with them just fine.

Alexander Skwar
-- 
Economists are still trying to figure out why the girls with the least
principle draw the most interest.
-- 
gentoo-user@g.o mailing list