Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

CocaineJohnsson · 2021-05-20 19:44:55

Prusa slicer links to libImath-3_0.so.27 and libImath-3_0.so.28, but extra/imath 3.0.3-1 only has libImath-3_0.so.28 resulting in the program being impossible to run.

I regenerated the ld cache using ldconfig, and updated locate's database (which is very slow) and neither can find it

$ locate libImath-3_0.so
/usr/lib/libImath-3_0.so
/usr/lib/libImath-3_0.so.28
/usr/lib/libImath-3_0.so.28.0.0

 ldd /usr/bin/prusa-slicer
	linux-vdso.so.1 (0x00007fff201d0000)

	libtbb.so.2 => /usr/lib/libtbb.so.2 (0x00007f60f2b64000)
	libImath-3_0.so.28 => /usr/lib/libImath-3_0.so.28 (0x00007f60f2b10000)

        /* I took the liberty to cut out some redundant diagnostic, we obviously don't care about all the working libraries ;) */

	libboost_iostreams.so.1.75.0 => /usr/lib/libboost_iostreams.so.1.75.0 (0x00007f60eef33000)
	libImath-3_0.so.27 => not found
	liblog4cplus-2.0.so.3 => /usr/lib/liblog4cplus-2.0.so.3 (0x00007f60eeea3000)

I can't locate a copy of libImath-3_0.so.27 using pacman -Fy implying it's not in any package, and I don't know if this qualifies as a package bug or not (nor which package the bug belongs to).

Why does it link to both .27 and .28, and if it's meant to why doesn't it provide the missing one, or imath provide both? Which package, if either, does this qualify as a bug for?

arojas · 2021-05-21 06:14:44

CocaineJohnsson wrote:

Why does it link to both .27 and .28, and if it's meant to why doesn't it provide the missing one, or imath provide both? Which package, if either, does this qualify as a bug for?

It doesn't. Your system is partially updated.

Slithery · 2021-05-21 10:20:40

arojas wrote:

It doesn't. Your system is partially updated.

It does according to the soname list on our website...
https://archlinux.org/packages/communit … r/sonames/

progandy · 2021-05-21 10:25:15

That soname list seems to be outdated, I installed prusa-slicer just now and ldd / lddtree only found a link to the newer .28. Maybe a dependency was updated after prusa-slicer itself and that inconsistent state was in the public mirrors for some time.
Edit: imath itself seems to have that strange soname list: https://archlinux.org/packages/extra/x86_64/imath/ (ldd on the package contents does not show anything wrong, though)

Last edited by progandy (2021-05-21 10:35:59)

CocaineJohnsson · 2021-05-21 14:16:04

arojas wrote:

It doesn't. Your system is partially updated.

I really don't see how that would make much sense since both prusa-slicer and imath are up to date, a quick pacman -Syu only reveals unrelated updates (nvidia driver, linux kernel 5.12.5, blender 2.92.0-7).

Maybe I misunderstand how ldd works, but in my understanding ldd reads the executable to find what shared objects it wants, and then prints that as a nice human readable output -- no? So if my prusa-slicer (2.3.1-2) and imath (3.0.3-1) packages are both fully up to date, then I don't really understand how a partially up to date system could be a problem here.

Rather, installing misc. updates and rebooting as a "solution" just feels like shaking a voodoo stick at the problem, if it solves it we learn nothing and if it doesn't solve it we gain nothing. But sure, I'll shake that voodoo stick -- though I'd like an explanation as to why it's expected to work.

V1del · 2021-05-21 14:19:55

run lddtree from pax-utils as opposed to ldd. It's potentially likely that some "other" binary has the actually wrong dependency here and ldd will just flatten the tree/dependency structure without giving you information into actual relation of things.

CocaineJohnsson · 2021-05-21 16:04:27

V1del wrote:

run lddtree from pax-utils as opposed to ldd. It's potentially likely that some "other" binary has the actually wrong dependency here and ldd will just flatten the tree/dependency structure without giving you information into actual relation of things.

... yeah that suddenly makes a lot more sense, given how many things use imath (though I still think this is suboptimal design in one way or another, why would ld think a software wants the wrong version of a library? Well whatever, that's a topic for another time and place)

arojas · 2021-05-21 16:30:15

Slithery wrote:

arojas wrote:
It doesn't. Your system is partially updated.
It does according to the soname list on our website...
https://archlinux.org/packages/communit … r/sonames/

That's a bug on the website. It is just not possible for a package to link to two different versions of the same library, simply because they can't both be simultaneously installed at build time.

Arch Linux

#1 2021-05-20 19:44:55

Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

#2 2021-05-21 06:14:44

Re: Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

#3 2021-05-21 10:20:40

Re: Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

#4 2021-05-21 10:25:15

Re: Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

#5 2021-05-21 14:16:04

Re: Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

#6 2021-05-21 14:19:55

Re: Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

#7 2021-05-21 16:04:27

Re: Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

#8 2021-05-21 16:30:15

Re: Prusa slicer links to 2 versions of libImath-3 only 1 in imath package

Board footer