[gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Maxim Kammerer]

Hi, virtual/libusb:0 has:

RDEPEND="|| ( >=dev-libs/libusb-compat-0.1.4
>=dev-libs/libusb-0.1.12-r7:0 >=sys-freebsd/freebsd-lib-8.0[usb] )"

However, after building a system from stage3, I still ended with
dev-libs/libusb:0 instead of dev-libs/libusb-compat (whereas stage3
has no libusb at all):

# equery d libusb
 * These packages depend on libusb:
app-crypt/ccid-1.4.8 (usb ? virtual/libusb:1)
app-crypt/gnupg-2.0.19 (usb ? virtual/libusb:0)
dev-libs/openobex-1.5 (usb ? virtual/libusb:0)
media-libs/libmtp-1.1.5 (virtual/libusb:1)
net-libs/libpcap-1.3.0-r1 (canusb ? virtual/libusb)
net-wireless/bluez-4.101-r5 (usb ? virtual/libusb:0)
sys-apps/pcsc-lite-1.8.6 (libusb ? virtual/libusb:1)
sys-apps/usb_modeswitch-1.2.5_p20121109 (virtual/libusb:0)
sys-apps/usbutils-006 (virtual/libusb:1)
virtual/libusb-0 (>=dev-libs/libusb-0.1.12-r7:0)
virtual/libusb-1 (>=dev-libs/libusb-1.0.9:1)

Any idea on what's going on? BFS instead of DFS search when satisfying "||"?

-- 
Maxim Kammerer
Liberté Linux: http://dee.su/liberte

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Peter Stuge]

Maxim Kammerer wrote:
>  * These packages depend on libusb:

One stands out:

> app-crypt/ccid-1.4.8 (usb ? virtual/libusb:1)
> app-crypt/gnupg-2.0.19 (usb ? virtual/libusb:0)
> dev-libs/openobex-1.5 (usb ? virtual/libusb:0)
> media-libs/libmtp-1.1.5 (virtual/libusb:1)
> net-libs/libpcap-1.3.0-r1 (canusb ? virtual/libusb)

This one has no slotted dependency. Does that matter? In any case it
doesn't seem completely correct, since the two APIs are not
compatible.


> Any idea on what's going on? BFS instead of DFS search when
> satisfying "||"?

Seems a good explanation.. Can you try swapping the two in the virtual?


//Peter

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Diego Elio Pettenò]

On 11/02/2013 16:17, Peter Stuge wrote:
>> > Any idea on what's going on? BFS instead of DFS search when
>> > satisfying "||"?
> Seems a good explanation.. Can you try swapping the two in the virtual?

Or not.

-- 
Diego Elio Pettenò — Flameeyes
flameeyes@flameeyes.eu — http://blog.flameeyes.eu/

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Zac Medico]

On 02/11/2013 06:18 AM, Maxim Kammerer wrote:
> Any idea on what's going on? BFS instead of DFS search when satisfying "||"?

It searches from left to right. If you can reproduce the problem, then
please create a debug logs as follows, and attach it to a bug on
bugs.gentoo.org:

   emerge [args] --pretend --debug > debug.log 2>&1
   xz -9 debug.log
-- 
Thanks,
Zac

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Maxim Kammerer]

On Mon, Feb 11, 2013 at 5:17 PM, Peter Stuge <peter@stuge.se> wrote:
>> net-libs/libpcap-1.3.0-r1 (canusb ? virtual/libusb)
>
> This one has no slotted dependency. Does that matter? In any case it
> doesn't seem completely correct, since the two APIs are not
> compatible.

It doesn't matter in this case, because canusb is disabled anyway. The
real dependencies are:

app-crypt/ccid-1.4.8 (usb ? virtual/libusb:1)
dev-libs/libusb-compat-0.1.4 (virtual/libusb:1)
dev-libs/openobex-1.5 (usb ? virtual/libusb:0)
media-libs/libmtp-1.1.5 (virtual/libusb:1)
net-wireless/bluez-4.101-r5 (usb ? virtual/libusb:0)
sys-apps/usb_modeswitch-1.2.5_p20121109 (virtual/libusb:0)
sys-apps/usbutils-006 (virtual/libusb:1)
virtual/libusb-0 (>=dev-libs/libusb-0.1.12-r7:0)
virtual/libusb-1 (>=dev-libs/libusb-1.0.9:1)

>> Any idea on what's going on? BFS instead of DFS search when
>> satisfying "||"?
>
> Seems a good explanation.. Can you try swapping the two in the virtual?

BFS and DFS both work left-to-right, but in absence of both
dev-libs/libusb:0 or :1, the path via libusb-compat to
dev-libs/libusb:1 is longer than the immediate path to
dev-libs/libusb:0. I think libusb-compat was selected correctly
previously (a few months ago), so perhaps and update in portage caused
the issue. On the other hand, this is just a hypothesis.

-- 
Maxim Kammerer
Liberté Linux: http://dee.su/liberte

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Zac Medico]

On 02/11/2013 07:42 AM, Maxim Kammerer wrote:
> On Mon, Feb 11, 2013 at 5:17 PM, Peter Stuge <peter@stuge.se> wrote:
>>> net-libs/libpcap-1.3.0-r1 (canusb ? virtual/libusb)
>>
>> This one has no slotted dependency. Does that matter? In any case it
>> doesn't seem completely correct, since the two APIs are not
>> compatible.
> 
> It doesn't matter in this case, because canusb is disabled anyway. The
> real dependencies are:
> 
> app-crypt/ccid-1.4.8 (usb ? virtual/libusb:1)
> dev-libs/libusb-compat-0.1.4 (virtual/libusb:1)
> dev-libs/openobex-1.5 (usb ? virtual/libusb:0)
> media-libs/libmtp-1.1.5 (virtual/libusb:1)
> net-wireless/bluez-4.101-r5 (usb ? virtual/libusb:0)
> sys-apps/usb_modeswitch-1.2.5_p20121109 (virtual/libusb:0)
> sys-apps/usbutils-006 (virtual/libusb:1)
> virtual/libusb-0 (>=dev-libs/libusb-0.1.12-r7:0)
> virtual/libusb-1 (>=dev-libs/libusb-1.0.9:1)
> 
>>> Any idea on what's going on? BFS instead of DFS search when
>>> satisfying "||"?
>>
>> Seems a good explanation.. Can you try swapping the two in the virtual?
> 
> BFS and DFS both work left-to-right, but in absence of both
> dev-libs/libusb:0 or :1, the path via libusb-compat to
> dev-libs/libusb:1 is longer than the immediate path to
> dev-libs/libusb:0. I think libusb-compat was selected correctly
> previously (a few months ago), so perhaps and update in portage caused
> the issue. On the other hand, this is just a hypothesis.

There has been no change in the portage behavior. Since libusb-compat is
on the left, it's supposed to be preferred.
-- 
Thanks,
Zac

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Maxim Kammerer]

On Mon, Feb 11, 2013 at 5:40 PM, Zac Medico <zmedico@gentoo.org> wrote:
> If you can reproduce the problem, then
> please create a debug logs as follows, and attach it to a bug on
> bugs.gentoo.org:

Can't reproduce with stage3 + emerge -upvDN. :/
I will look out for this issue in full system builds in the future.

-- 
Maxim Kammerer
Liberté Linux: http://dee.su/liberte

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Peter Stuge]

Diego Elio Pettenò wrote:
> > Can you try swapping the two in the virtual?
> 
> Or not.

I guess that you assumed that I suggested to test this in-tree, so
I guess I should clarify that I would expect it to be tested in
PORTDIR_OVERLAY.

If my guess is correct then you are really way too eager to
misunderstand what people intend to transmit, given a less than
unambiguous message.

In the future, please give me (and why not others) the benefit of the
doubt, and ask for clarification if I suggest something that seems
like a bad idea to you.

It does require more email round trips, but the end result is not
only a positive discussion which promotes greater understanding for
active and passive participants alike - it also makes you seem much
more sympathetic which can only be a good thing.


//Peter

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Diego Elio Pettenò]

On 11/02/2013 17:44, Peter Stuge wrote:
> If my guess is correct then you are really way too eager to
> misunderstand what people intend to transmit, given a less than
> unambiguous message.

No, it's because of what Maxim already said: it's not an LTR/RTL issue.

-- 
Diego Elio Pettenò — Flameeyes
flameeyes@flameeyes.eu — http://blog.flameeyes.eu/

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Peter Stuge]

Diego Elio Pettenò wrote:
> it's because of what Maxim already said: it's not an LTR/RTL issue.

Do you have an idea about what the issue is?


//Peter

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Diego Elio Pettenò]

On 11/02/2013 17:54, Peter Stuge wrote:
> Do you have an idea about what the issue is?

No, but I'm pretty certain that it's not that, because the preference is
and has been for a very long time LTR.

Which happens to be one of the things the quizzes are there to ensure
people know.

-- 
Diego Elio Pettenò — Flameeyes
flameeyes@flameeyes.eu — http://blog.flameeyes.eu/

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Zac Medico]

On 02/11/2013 08:54 AM, Peter Stuge wrote:
> Diego Elio Pettenò wrote:
>> it's because of what Maxim already said: it's not an LTR/RTL issue.
> 
> Do you have an idea about what the issue is?

My guess is that there were one or more ebuilds that inappropriately
specified dev-libs/libusb:0 instead of virtual/libusb:0, and they have
since been fixed.
-- 
Thanks,
Zac

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Peter Stuge]

Zac Medico wrote:
> My guess is that there were one or more ebuilds that inappropriately
> specified dev-libs/libusb:0 instead of virtual/libusb:0, and they have
> since been fixed.

I believe they were all changed some months ago, but it's of course still
possible if either the snapshot was old or if a new ebuild uses dev-libs.


//Peter

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Maxim Kammerer]

On Mon, Feb 11, 2013 at 7:09 PM, Zac Medico <zmedico@gentoo.org> wrote:
> My guess is that there were one or more ebuilds that inappropriately
> specified dev-libs/libusb:0 instead of virtual/libusb:0, and they have
> since been fixed.

I did the full build yesterday, with most recent stage3 and portage
snapshot. All the involved ebuilds are older than the snapshot. I will
try another full build tomorrow, and if the issue persists, try to
uncover the culprit (naturally, my build script is somewhat more
complex than emerge -upvDN @world I tried above, so hopefully the
issue is reproducible).

-- 
Maxim Kammerer
Liberté Linux: http://dee.su/liberte

Re: [gentoo-dev] libusb-compat preference in virtual/libusb:0 not strong enough?

[Maxim Kammerer]

On Mon, Feb 11, 2013 at 8:05 PM, Maxim Kammerer <mk@dee.su> wrote:
> I will try another full build tomorrow, and if the issue persists, try to
> uncover the culprit (naturally, my build script is somewhat more
> complex than emerge -upvDN @world I tried above, so hopefully the
> issue is reproducible).

Not reproducible this time:

# emerge -upvDN --with-bdeps y @world | grep /libusb
[ebuild  N     ] dev-libs/libusb-1.0.9:1  USE="-debug -doc -static-libs" 413 kB
[ebuild  N     ] virtual/libusb-1:1  0 kB
[ebuild  N     ] dev-libs/libusb-compat-0.1.4  USE="-debug -static-libs" 237 kB
[ebuild  N     ] virtual/libusb-0  0 kB

Oh well.

-- 
Maxim Kammerer
Liberté Linux: http://dee.su/liberte