ChanServ changed the topic of #wayland to: https://wayland.freedesktop.org | Discussion about the Wayland protocol and its implementations, plus libinput | register your nick to speak
agd5f has quit [Read error: No route to host]
agd5f has joined #wayland
Hypfer is now known as Guest738
Hypfer has joined #wayland
Guest738 has quit [Read error: Connection reset by peer]
smallville7123 has joined #wayland
ybogdano has quit [Ping timeout: 480 seconds]
co1umbarius has joined #wayland
columbarius has quit [Ping timeout: 480 seconds]
Hypfer is now known as Guest744
Hypfer has joined #wayland
Guest744 has quit [Ping timeout: 480 seconds]
smallville7123 has quit [Ping timeout: 480 seconds]
molinari has quit [Ping timeout: 480 seconds]
ellemoeiselleva has joined #wayland
ellemoe_ has joined #wayland
ellemoeiselleva has quit [Ping timeout: 480 seconds]
Hypfer has quit [Ping timeout: 480 seconds]
Hypfer has joined #wayland
mxz has quit [Quit: cya]
mxz has joined #wayland
Hypfer is now known as Guest756
Hypfer has joined #wayland
Guest756 has quit [Ping timeout: 480 seconds]
<mjt>
daniels: adding major version to the lib is common in between the ones who has no clue on what is a soname :)
<mjt>
daniels: or when the abi isn't stable "enough" yet to declare it as such
<mjt>
"Never try to set the interface numbers so that they correspond to the release number of your package. This is an abuse that only fosters misunderstanding of the purpose of library versions."
ellemoe_ has quit []
nerdopolis has quit [Remote host closed the connection]
nerdopolis has joined #wayland
hardening has joined #wayland
Company has quit [Read error: Connection reset by peer]
kts has joined #wayland
kts has quit [Read error: Connection reset by peer]
<mjt>
what's needed these days to run (wlroots-based) compositors from a vt? I used to run just `weston' or `sway' from a text tty on linux. Now stuff errors out, -
<mjt>
Could not activate session: permission denied
<mjt>
ah, it's just `cage', sway and weston still works
<mjt>
any1: for fun (re that archlinux wiki link about sway#Virtualization), qemu -vga qxl has its own share of issues, - bugs in qemu with mouse cursor handling to begin with. It works find with -vga std but breaks badly with -vga qxl :))
<daniels>
mjt: sure, but there are plenty of packages out there who suggest that GNU are just noisily opinionated rather than objectively correct
<mjt>
it's not about gnu. it's about practical usage. if your abi didn't change in incompatible ways, there's no need to recompile all users of your library
kts has joined #wayland
<mjt>
when you bump your soname in every release, all users need to be recompiled
<mjt>
wenever you change the ABI in incompatible way or not, do you care to keep it if possible, etc, is an entirely different matter
<mjt>
you can declare that this library doesn't have stable ABI. Bumping soname in every release basically says, "don't use this library (yet)"
<mjt>
basically, with sonames like this, it becomes impractical to have this library packaged in a downstream distribution (redhat/debian/arch/suse/whatever)
jmdaemon has quit [Ping timeout: 480 seconds]
molinari has joined #wayland
mvlad has joined #wayland
molinari has quit [Ping timeout: 480 seconds]
molinari has joined #wayland
<daniels>
I know what an ABI change means, yeah
<daniels>
and it’s self-evidently not impractical, because all distributions _do_ package plenty of libs which encode a major into the name component to allow for parallel installation
<mjt>
it depends on how often these versions changes, basically :)
rasterman has joined #wayland
<mjt>
f.e. in debian one have to complete the "NEW package" procedure each time the soname changes, - which might require from several days to months, as each new package is being reviewed
<mjt>
(one can argue this is a limitation of debian though)
<daniels>
yeah, NEW isn’t exactly great
<daniels>
but I take the point about wanting to limit ABI changes and it is something we’re working on
molinari has quit [Ping timeout: 480 seconds]
<mjt>
it's not the abi changes, it's the *incompatible* abi changes
<mjt>
if the soname changes anyway, there's no reason to think about keeping ABI compatible
<daniels>
every ABI change is incompatible …
<daniels>
unless you mean additions only
<mjt>
nope, not at all. Say, a new value for an existing arument of a function
rv1sr has joined #wayland
<mjt>
if you need to add another argument to a function, you can just do it and break ABI, or you can introduce a new function with extra argument and keep existing function (and call the new functioon from the existing)
<mjt>
(depending on common sense, - it's not a good idea to have tons of old compat functions too)
Dami_Lu has quit [Remote host closed the connection]
caveman has quit [Remote host closed the connection]
<pq>
mjt, yes, we know very well of soname and libtool versioning. That's exactly why we chose not to use them. The GNU library style is far too fragile and confusing. The very reason we do this is to make libweston packageable when we do break ABI every release, because there is no designed API yet. See "parallel installability" in Weston readme.
kts has quit [Remote host closed the connection]
<mjt>
pq: I guessed exactly as much: see my comment above, "or when the interface isn't stable (yet)"
<pq>
mjt, and yes, we do know how we can avoid ABI breaks. However, the current API makes it quite impossible for many changes, because we expose far too much of the internal details, and cannot simply just close it all off in one go. People actually like the exposed internals, unfortunately.
<mjt>
yeah, that definitely makes sense
<mjt>
exposing internals is a two-edge sword too
<pq>
so it was a very deliberate decisions to put major version in lib name, and not use the hidden soname versioning.
<pq>
I don't appreciate being implied that we are just stupid.
<mjt>
I come with a different background. libiscsi for example (I maintain(ed) it in debian): sometimes a new "major" release includes a few bugfixes (with no abi/api change whatsoever) and a new example binary, and the soname bump to match the release number. They did say their abi is stable though. At one point I gave up and start patching. So far, debian has libiscsi7, upsrteam is at 19
molinari has joined #wayland
<mjt>
my comment was a reply to daniels (not implying anyone's stupid or the like) about gnu
<mjt>
I tend to evaluate actual technical stuff each time. not everything gnu is bad
Company has joined #wayland
<mjt>
or a joke (which come before). either way, it was definitely friendly smile, not intended to punch anyone