[Pkg-samba-maint] Bug#974868: samba-vfs-modules: Still causing issues - at least on armv5tel/armel

[Michael Tokarev]

Control: retitle -1 samba: can't serve size-limited Time Machine shares on 32bit architectures
Control: tag -1 + confirmed upstream
Control: severity -1 minor

16.11.2022 21:29, ArtMG wrote:
> Hi,
> 
>  > On Mon, 31 Oct 2022 14:19:24 +0300 Michael Tokarev <mjt at tls.msk.ru <mailto:mjt at tls.msk.ru>> wrote:
>  > > Lacking any further information, I'll close this bugreport as fixed in 4.16.
> 
> I have just built and tested against 4.16.7 and I'm afraid to say that the reported issue *DOES* still occur.

Ok.

It's good I didn't actually close this bug report as I wanted to - I thought
about sending the email to nnn-done at bugs.d.o, but forgot about this -done part.
(I was dealing with old bugs, there are *lots* of them reported against samba
package, they're just sitting there for years without any action or attention,
so something has to be done with them anyway).


> On Tue, 15 Nov 2022 11:03:02 +0300 Michael Tokarev <mjt at tls.msk.ru <mailto:mjt at tls.msk.ru>> wrote:
>  > If you think this is incorrect and the issue is still here with current version
>  > of samba, please feel free to reopen this bug report with any extra information
>  > which might be helpful to identify the issue.
> 
> What kind of extra information might be helpful? I can pull diagnostics off my test rig if it helps.

I for one don't know what vfs_fruit *is*, to begin with. Just read briefly the
manpage, -- well, it has quite some things, it seems, most of which is related
to MacOS.  I don't have a MacOS machine.

So basically, I know nothing about how to verify this.

And the thing is that I can't really do anything there.
It's best to be handled upstream anyway, I don't know
samba internals.  I can handle packaging issues, but
for the real code issues, especially the ones which I
can't even verify myself - I can only close the bug
reports so the reports wont stack and hide somethin
which I really can fix.

Can we at least retitle this as "... on 32bi platforms",
since what you describe suggest that it's not specific
to 32bits.

> Part of me wants to get this issue resolved so that 32-bit systems can meet these use-cases.
> However, I know from working on the patch for that other 
> bug (https://bugzilla.samba.org/show_bug.cgi?id=13622#c10) that you increase the precision in 
> one part of the code, and sooner or later an issue arises in another.

Umm. So we've hit some internal limitations with off_t size
here.  That's understandable. Just yesterday we talked with
Qemu folks - they just *detest* 32bit platforms, since it's
extremly difficult to map even the 32bit address space into
its own 32bit address space, they have to perform all kinds
of tricks which doesn't work anyway in the end, - it is just
can not be done.

> In the end, there's a limit how long it's pragmatic to play bug whack-a-mole like this.
>  Since my previous issue I have upgraded to hardware that 
> supports 64-bit, so I'm now off to validate whether your suggestion of upgrading
> OS architecture can make this issue magically disappear.

I see. Well, this is at least practical, I think.

Thank you very much for your input!

/mjt