Biggest FreeBSD update?

[RetroComputingCollector]

What do you think the biggest major FreeBSD version release was, in terms of the amount of users its changes would impact (not counting dropped hardware), and the amount of changes it made? I am curious to see what people's experiences with prior updates were.

 

[covacat]

4 to 5

 

[cracauer@]

4 to 5

Yes.

 

[covacat]

I admit I chickened out and skipped 5.
4.x was my all time favorite version.

 

[vmisev]

I agree with covacat - 4 to 5 was biggest (as I heard, I started with 6), but also important changes in 10 (switch from gcc to clang) and 13 (switch from svn to git, OpenZFS), and now upcoming 15 with pkgbase

 

[vmisev]

4.x was my all time favorite version.

I reckon that Mat Dillon will agree with you on that

 

[mer]

When did we switch from a.out to dynamic libs?
4 to 5 was a big change (birthing DragonFlyBSD). It took a little longer for 5.x to stabilize, which turned some people away, but I think looking back 5 was a good thing.

For me, I'm not sure how big gcc to clang was: Important, yes, absolutely, but not something a pure user sees.
svn to git: I think the cvs to svn was a bigger change but choice of SCM platform typically only affects people contributing or building from source.

 

[covacat]

a.out to elf was 2.2.something iirc

 

[mer]

a.out to elf was 2.2.something iirc

Ok, so the "first" release post lawsuits were 3.something and already dynamic. Too much history, too much time.

 

[vmisev]

For me, I'm not sure how big gcc to clang was: Important, yes, absolutely, but not something a pure user sees.
svn to git: I think the cvs to svn was a bigger change but choice of SCM platform typically only affects people contributing or building from source.

For people running STABLE () both changes were if not important, then something one has to get used to. I still know jack about git except how to pull.

I already forgot about cvs When it was that we switched to svn? 9.1, 9.2 somewhere there...

 

[covacat]

a.out version had shared libs too. just the file format was not elf but a.out

I always run -stable back then, binary upgrades were not supported.
but csup/buildworld was reasonably fast even on shitty hardware

 

[vmisev]

a.out to elf was 2.2.something iirc

3.0

 

[SirDice]

Same here, 4 to 5. Messed up the upgrade, significantly. Didn't read /usr/src/UPDATING, the switch to UFS2 meant you needed install the kernel and MUST reboot to activate the new 5.0 kernel before installing world. Only read those instructions after the whole thing fell apart

Other than that I really had no issues with 5.0, at least nothing significant enough to remember almost 20 years later.

 

[cracauer@]

My most unfortunate memory of FreeBSD versions is a period when X11 would come up correctly but after a while of idling would lose keyboard and mouse/trackpad. Had that with several laptops. Ca. 2019.

 

[mer]

Same here, 4 to 5. Messed up the upgrade, significantly. Didn't read /usr/src/UPDATING, the switch to UFS2 meant you needed install the kernel and MUST reboot to activate the new 5.0 kernel before installing world. Only read those instructions after the whole thing fell apart

Other than that I really had no issues with 5.0, at least nothing significant enough to remember almost 20 years later.

I think that bit a few people and of course RTFM BEFORE doing an upgrade? That's crazy talk. If I remember correctly seeing a few of those on mailing lists is what prompted me to start doing the "get a new device and fresh install" routine until ZFS came along.

 

[covacat]

4.x had (by default) a smallish rootfs and everything on / was statically linked (/usr was a separate fs)
there was no /lib
i know i updated some boxes from 4.x to 6.x remotely (without ipmi / ipkvm) and it was awkward
4.x root / was too small to fit the 6.x / so i did some acrobatics with the partitions like patching /etc/rc to
shrink swap and growfs / before mounting / rw. i simulated the stuff before because most of the boxes were hundreds of km away.

 

[vmisev]

4.x had (by default) a smallish rootfs and everything on / was statically linked (/usr was a separate fs)
there was no /lib
i know i updated some boxes from 4.x to 6.x remotely (without ipmi / ipkvm) and it was awkward
4.x root / was too small to fit the 6.x / so i did some acrobatics with the partitions like patching /etc/rc to
shrink swap and growfs / before mounting / rw. i simulated the stuff before because most of the boxes were hundreds of km away.

This reminded me of one time when I had to enlarge and shuffle partitions because new loader.efi was too big for ESP, can't remember was it 10 to 11 or 11 to 12?