Mailing List Archive

On 4/8/24 07:03, Dr Rainer Woitok wrote:
> Greetings,
>
> the upgrade on my old laptop with two 2.7GHz Dual-Core Skylake proces-
> sors took slightly more than 2 hours for the manual upgrading of "bin-
> utils", "gcc" and "glibc", and slightly more than 21.5 hours for the fi-
> nal upgrade of "@world", which had to process a total of 1061 packages.
> I'm wondering whether a fresh install from a stage 3 "tar" ball would
> have been faster?
>
> My first Gentoo installation on this laptop back in mid 2019 used pro-
> file 17.1 (which is still marked "experimental", by the way). Now, less
> than five years later this profile set is deprecated. Is five years a
> common intervall between enforced Gentoo profile upgrades?
>
> Sincerely,
> Rainer
>

I had to upgrade about 7 machines, and three wound up having weird
troubles - so I did exactly that and started fresh on the rest. Working
on the last one (my laptop) right now.

Dan

On 4/8/24 10:03 AM, Dr Rainer Woitok wrote:
> Greetings,
>
> the upgrade on my old laptop with two 2.7GHz Dual-Core Skylake proces-
> sors took slightly more than 2 hours for the manual upgrading of "bin-
> utils", "gcc" and "glibc", and slightly more than 21.5 hours for the fi-
> nal upgrade of "@world", which had to process a total of 1061 packages.
> I'm wondering whether a fresh install from a stage 3 "tar" ball would
> have been faster?


If you're okay doing a fresh install from a stage3 tar, which is faster
at least to install the base system because it is all precompiled and
you are not building the packages yourself, then I would assume you're
also okay doing the update using the gentoo.org official binhost.

They're both just the binaries that Gentoo's release automation builds
for you. Extracting a bunch of gpkgs is much faster than compiling them,
and not too much slower than extracting a single stage3 tarball.

It also has the advantage that for amd64, more than just the stage3
package set can be sped up like this -- and you don't have to rebuild
the installation, recreate @world, backup and restore user data, etc.

Just enable the binhost and then do the same -e @world you were doing
without the binhost. :)


> My first Gentoo installation on this laptop back in mid 2019 used pro-
> file 17.1 (which is still marked "experimental", by the way). Now, less
> than five years later this profile set is deprecated. Is five years a
> common intervall between enforced Gentoo profile upgrades?


Well, 13.0 -> 17.0 -> 17.1 -> 23.0 so I suppose you could say they are
fairly long intervals, yeah.

As far as it being marked experimental: it was dropped from stable
during the 23.0 announcement, but is being marked as stable again:

https://github.com/gentoo/gentoo/pull/35871

Rationale:

"""
Making 17.1 exp immediately gives the impression that it's formally
deprecated, which it isn't yet.
"""


--
Eli Schwartz

I use a buildhost for each of the 4 architectures I manage - binary
emtytree installs are not to bad.  However the initial build for low
power arm systems is measured in multiple days (for just the initial
toolchain, not hours :(.  Only minor problems so far though which is
good.  At least it can build while online, unlike fresh installs which
mean lots of downtime and more work for me in configuring.

BillK


On 9/4/24 05:14, Eli Schwartz wrote:
> On 4/8/24 10:03 AM, Dr Rainer Woitok wrote:
>> Greetings,
>>
>> the upgrade on my old laptop with two 2.7GHz Dual-Core Skylake proces-
>> sors took slightly more than 2 hours for the manual upgrading of "bin-
>> utils", "gcc" and "glibc", and slightly more than 21.5 hours for the fi-
>> nal upgrade of "@world", which had to process a total of 1061 packages.
>> I'm wondering whether a fresh install from a stage 3 "tar" ball would
>> have been faster?
>
> If you're okay doing a fresh install from a stage3 tar, which is faster
> at least to install the base system because it is all precompiled and
> you are not building the packages yourself, then I would assume you're
> also okay doing the update using the gentoo.org official binhost.
>
> They're both just the binaries that Gentoo's release automation builds
> for you. Extracting a bunch of gpkgs is much faster than compiling them,
> and not too much slower than extracting a single stage3 tarball.
>
> It also has the advantage that for amd64, more than just the stage3
> package set can be sped up like this -- and you don't have to rebuild
> the installation, recreate @world, backup and restore user data, etc.
>
> Just enable the binhost and then do the same -e @world you were doing
> without the binhost. :)
>
>
>> My first Gentoo installation on this laptop back in mid 2019 used pro-
>> file 17.1 (which is still marked "experimental", by the way). Now, less
>> than five years later this profile set is deprecated. Is five years a
>> common intervall between enforced Gentoo profile upgrades?
>
> Well, 13.0 -> 17.0 -> 17.1 -> 23.0 so I suppose you could say they are
> fairly long intervals, yeah.
>
> As far as it being marked experimental: it was dropped from stable
> during the 23.0 announcement, but is being marked as stable again:
>
> https://github.com/gentoo/gentoo/pull/35871
>
> Rationale:
>
> """
> Making 17.1 exp immediately gives the impression that it's formally
> deprecated, which it isn't yet.
> """
>
>

On Monday, 8 April 2024 22:14:30 BST Eli Schwartz wrote:

> If you're okay doing a fresh install from a stage3 tar, which is faster
> at least to install the base system because it is all precompiled and
> you are not building the packages yourself, then I would assume you're
> also okay doing the update using the gentoo.org official binhost.
>
> They're both just the binaries that Gentoo's release automation builds
> for you. Extracting a bunch of gpkgs is much faster than compiling them,
> and not too much slower than extracting a single stage3 tarball.
>
> It also has the advantage that for amd64, more than just the stage3
> package set can be sped up like this -- and you don't have to rebuild
> the installation, recreate @world, backup and restore user data, etc.
>
> Just enable the binhost and then do the same -e @world you were doing
> without the binhost. :)

There is one caveat, though: all the binary packages have been compiled with
default USE flags. If you've changed any on your system, you'll still have to
install those packages the standard way. I have 24 such USE settings on this
machine.

--
Regards,
Peter.

On 08/04/2024 15:03, Dr Rainer Woitok wrote:
> the upgrade on my old laptop with two 2.7GHz Dual-Core Skylake proces-
> sors took slightly more than 2 hours for the manual upgrading of "bin-
> utils", "gcc" and "glibc", and slightly more than 21.5 hours for the fi-
> nal upgrade of "@world", which had to process a total of 1061 packages.
> I'm wondering whether a fresh install from a stage 3 "tar" ball would
> have been faster?

Some 1500 plus packages here - took about 2 days on my 4-core Ryzen ...

Btw, where are all the messages for packages stored? I ought to go
through them and make sure there aren't any messages of interest... I
know I ought to update my kernel ...

Cheers,
Wol

On 4/9/24 5:55 AM, Peter Humphrey wrote:
> There is one caveat, though: all the binary packages have been compiled with
> default USE flags. If you've changed any on your system, you'll still have to
> install those packages the standard way. I have 24 such USE settings on this
> machine.


But within the context under discussion, reinstalling from a stage3 and
then modifying those USE flags and re-emerging would *also* require
installing those packages the standard way.

(Note that for amd64 and arm64, you can actually get multiple USE
variants, since it builds packages for server, gnome and kde profiles.)


--
Eli Schwartz

Wol,

On Tuesday, 2024-04-09 18:36:53 +0100, you wrote:

> ...
> Btw, where are all the messages for packages stored? I ought to go
> through them and make sure there aren't any messages of interest...

My script for package installations or upgrades sets

begin=$(date '+%Y-%m-%d %H:%M:%S %Z')

before it calls "emerge" and calls the following little "gawk" programme
after "emerge" has finished:

gawk -v begin="$begin" '
! P && /^>>> M/ { match($0," on ([^ ]+ [^ ]+ [^ ]+) for ",m)
if ( m[1] < begin ) next # Skip old messages.
printf "\n" # Print separator before first message.
P = 1 # Print remaining messages.
}
P ' /var/log/portage/elog/summary.log | more

However, this probably requires

PORTAGE_ELOG_SYSTEM="save-summary:warn"
PORTAGE_LOGDIR="/var/log/portage"

in your "make.conf" file.

Sincerely,
Rainer