View previous topic :: View next topic |
Author |
Message |
pjp Administrator
Joined: 16 Apr 2002 Posts: 20067
|
Posted: Sat Apr 06, 2024 1:30 am Post subject: [solved] Remaining life of profile 17.1? |
|
|
With a cursory look at the benefits of 23, my hardware is too old, at least for cet.
Per the euse info for cet on gcc, the benefits require at least Tiger Lake or Zen 3. I have a Phenom and a Kaby Lake. Not being harmful isn't much of a motivation for --emptytree. Similar for -fcf-protection with hardened.
Is 17.1 likely to be EOL sooner than later, or is it likely to be around for multiple years? _________________ Quis separabit? Quo animo?
Last edited by pjp on Sat Apr 06, 2024 6:35 am; edited 1 time in total |
|
Back to top |
|
|
grknight Retired Dev
Joined: 20 Feb 2015 Posts: 1702
|
Posted: Sat Apr 06, 2024 1:36 am Post subject: |
|
|
That's not the only thing that 23.0 does.. 2 other options on sys-devel/gcc become forced:
Code: | F F default-stack-clash-protection : Build packages with stack clash protection on by default as a hardening measure. This
enables -fstack-clash-protection by default which protects against large memory
allocations allowing stack smashing. May cause slightly increased codesize, but modern
compilers have been adapted to optimize well for this case, as this mitigation is now
quite common. See
https://developers.redhat.com/blog/2020/05/22/stack-clash-mitigation-in-gcc-part-3 and
https://www.qualys.com/2017/06/19/stack-clash/stack-clash.txt.
F F default-znow : Request full relocation on start from ld.so by default. This sets the -z,now (BIND_NOW)
flag by default on all linker invocations. By resolving all dynamic symbols at application
startup, parts of the program can be made read-only as a hardening measure. This is
closely related to RELRO which is also separately enabled by default. In some applications
with many unresolved symbols (heavily plugin based, for example), startup time may be
impacted. |
Usually, older profiles exist for 6-12 months before removal, but I'm not sure of any official plans. There is usually an announcement before they do go. |
|
Back to top |
|
|
pjp Administrator
Joined: 16 Apr 2002 Posts: 20067
|
Posted: Sat Apr 06, 2024 1:55 am Post subject: |
|
|
Thanks. I presumed it was too early, but thought I'd check. _________________ Quis separabit? Quo animo? |
|
Back to top |
|
|
Phoenix591 Guru
Joined: 17 Sep 2007 Posts: 488
|
|
Back to top |
|
|
figueroa Advocate
Joined: 14 Aug 2005 Posts: 2966 Location: Edge of marsh USA
|
Posted: Sat Apr 06, 2024 3:09 am Post subject: |
|
|
I had my Phenom x3 on "[5] default/linux/x86/17.0/desktop (exp)." Updating to "[28] default/linux/x86/23.0/i686/split-usr/desktop (stable)" posed no issue. The --emptytree ran for about 24 hours, but the machine is on 24/7 anyway. That machine has 8 GB RAM.
This particular machine has been running since 2012. It seemed like a good idea to avoid going obsolete. _________________ Andy Figueroa
hp pavilion hpe h8-1260t/2AB5; spinning rust x3
i7-2600 @ 3.40GHz; 16 gb; Radeon HD 7570
amd64/23.0/split-usr/desktop (stable), OpenRC, -systemd -pulseaudio -uefi |
|
Back to top |
|
|
pjp Administrator
Joined: 16 Apr 2002 Posts: 20067
|
Posted: Sat Apr 06, 2024 6:35 am Post subject: |
|
|
@Phoenix591,
Thanks! That gives me a deadline to come up with a solution of some kind.
@figueroa,
I have some sort of issue that I presume is hardware related, maybe overheating. I've had larger packages primarily gcc, clang fail, usually with success on the subsequent attempt. *knock on wood* I haven't seen it in a while. I may not replace the hardware, so I have no replacement plans at this time. That would leave me with laptops, and effectively no system with which to build binaries.
I may try to find a way to build packages in groups, but that seems complicated. So I may end up riding out 17.1 until the end. _________________ Quis separabit? Quo animo? |
|
Back to top |
|
|
NichtDerHans Tux's lil' helper
Joined: 27 Jan 2023 Posts: 138
|
Posted: Sat Apr 06, 2024 10:09 am Post subject: |
|
|
I can recommend to do the compiling via chroot on a faster computer. I keep my X220 laptop up to date every two weeks. It works very well.
Export the filesystem from the slow PC via NFS in the local Network, mount on the faster PC, chroot to the mountpoint. |
|
Back to top |
|
|
NeddySeagoon Administrator
Joined: 05 Jul 2003 Posts: 54397 Location: 56N 3W
|
Posted: Sat Apr 06, 2024 10:18 am Post subject: |
|
|
pjp,
If you want to stay on /17.x/ copy the profile to your overlay. Use that instead of the profile from ::gentoo.
Now it gets a bit murky. How much of the profile do you want to keep?
Its possible to copy just parts of a profile too. _________________ Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
Back to top |
|
|
Anon-E-moose Watchman
Joined: 23 May 2008 Posts: 6103 Location: Dallas area
|
Posted: Sat Apr 06, 2024 11:12 am Post subject: |
|
|
The easy way is to copy the whole /usr/portage/profile to your local repo, then do eselect profile and point to your local repo.
It only takes up ~3 meg disk space.
I did the whole copy thing back in the 17.0 -> 17.1 days. No problems. _________________ PRIME x570-pro, 3700x, 6.1 zen kernel
gcc 13, profile 17.0 (custom bare multilib), openrc, wayland |
|
Back to top |
|
|
pjp Administrator
Joined: 16 Apr 2002 Posts: 20067
|
Posted: Sat Apr 06, 2024 3:18 pm Post subject: |
|
|
@NichtDerHans,
That is the "faster" computer :)
@NeddySeagoon, Anon-E-moose,
The initial concern there is what I'd be getting myself into regarding a system that is supported / supportable. As far as what to keep, I haven't the foggiest idea what is really in a profile. So having a Franken-profile doesn't seem like the best approach for Future Me... never mind how often I ignore that concern :).
A complication is that I find the structure of profiles to be an incomprehensible abyss. I'm sure some people understand how they work. _________________ Quis separabit? Quo animo? |
|
Back to top |
|
|
Anon-E-moose Watchman
Joined: 23 May 2008 Posts: 6103 Location: Dallas area
|
Posted: Sat Apr 06, 2024 10:27 pm Post subject: |
|
|
The profiles are indeed a little .. abstruse. The parent files are what's important.
Lets say your profile points to default/linux/amd64/17.1/desktop
in that directory, is an eapi file and a parent file.
Code: | $ cat parent
..
../../../../../targets/desktop |
So 2 things are referenced, the directory directly above which contains
Code: | $ cat parent
..
../../../../arch/amd64
../../../../releases/17.0 |
And the /targets directory contains make.* and package.* files, but no parent, so end of the line in that directlon
but you just keep on until there are no more parents to traverse.
It's just a matter of chasing down parents and applying any files found, make.defaults, package.*, etc _________________ PRIME x570-pro, 3700x, 6.1 zen kernel
gcc 13, profile 17.0 (custom bare multilib), openrc, wayland |
|
Back to top |
|
|
pjp Administrator
Joined: 16 Apr 2002 Posts: 20067
|
Posted: Sun Apr 07, 2024 4:00 am Post subject: |
|
|
I'll keep it in mind. Without having closely read the announcement, I'm leaning toward the binpkg method and a "manual" emptytree method. If I can deduce the "big" packages and exclude those, then maybe the emptytree option would work. Maybe mesa would be on the exclude list too? Other than gcc and llvm, nothing jumps out as obvious. Apparently I've set gdb to not use my tmpfs.
Then there's my chroot -> binpkg -> laptop situation. _________________ Quis separabit? Quo animo? |
|
Back to top |
|
|
|