YAMAMOTO Takashi | 12 Jun 2012 03:58
Picon

liblzf incompatibility

hi,

is there a near-future plan to use liblzf for any in-tree stuff?
otherwise i'd suggest to disable it because it only causes problems.
see PR/46426.

YAMAMOTO Takashi

Thor Lancelot Simon | 12 Jun 2012 05:45
Picon
Favicon

Re: liblzf incompatibility

On Tue, Jun 12, 2012 at 01:58:12AM +0000, YAMAMOTO Takashi wrote:
> hi,
> 
> is there a near-future plan to use liblzf for any in-tree stuff?
> otherwise i'd suggest to disable it because it only causes problems.
> see PR/46426.

I have code that uses it, both in userspace and the kernel, but I never
checked it in because I ran into a build problem with src/common.  Let me
see what I can do about both issues.

--

-- 
  Thor Lancelot Simon	                                     tls <at> panix.com

  "We cannot usually in social life pursue a single value or a single moral
   aim, untroubled by the need to compromise with others."      - H.L.A. Hart

Thor Lancelot Simon | 12 Jun 2012 18:12
Picon
Favicon

Re: liblzf incompatibility

On Mon, Jun 11, 2012 at 11:45:11PM -0400, Thor Lancelot Simon wrote:
> On Tue, Jun 12, 2012 at 01:58:12AM +0000, YAMAMOTO Takashi wrote:
> > hi,
> > 
> > is there a near-future plan to use liblzf for any in-tree stuff?
> > otherwise i'd suggest to disable it because it only causes problems.
> > see PR/46426.
> 
> I have code that uses it, both in userspace and the kernel, but I never
> checked it in because I ran into a build problem with src/common.  Let me
> see what I can do about both issues.

I am looking at the PR, and I don't really think I agree that our liblzf
is "incompatible with the upstream version".

Unfortunately, the upstream version can be compiled with either of two
different APIs.  I think the bug is actually in programs whose autoconf
logic detects liblzf without detecting which API is in use.

I chose the API that was more general, which I do think was the correct
decision.

Thor

YAMAMOTO Takashi | 12 Jun 2012 19:51
Picon

Re: liblzf incompatibility

hi,

> On Mon, Jun 11, 2012 at 11:45:11PM -0400, Thor Lancelot Simon wrote:
>> On Tue, Jun 12, 2012 at 01:58:12AM +0000, YAMAMOTO Takashi wrote:
>> > hi,
>> > 
>> > is there a near-future plan to use liblzf for any in-tree stuff?
>> > otherwise i'd suggest to disable it because it only causes problems.
>> > see PR/46426.
>> 
>> I have code that uses it, both in userspace and the kernel, but I never
>> checked it in because I ran into a build problem with src/common.  Let me
>> see what I can do about both issues.
> 
> I am looking at the PR, and I don't really think I agree that our liblzf
> is "incompatible with the upstream version".
> 
> Unfortunately, the upstream version can be compiled with either of two
> different APIs.  I think the bug is actually in programs whose autoconf
> logic detects liblzf without detecting which API is in use.

as far as i know, there's no reasonable way to detect the compile-time
option.  so assuming the default is reasonable.

> 
> I chose the API that was more general, which I do think was the correct
> decision.

i think the option is for users who embed the library into their
applications, not for general purpose OSes like us.
(Continue reading)

YAMAMOTO Takashi | 30 Jul 2012 02:14
Picon

Re: liblzf incompatibility

hi,

> hi,
> 
>> On Mon, Jun 11, 2012 at 11:45:11PM -0400, Thor Lancelot Simon wrote:
>>> On Tue, Jun 12, 2012 at 01:58:12AM +0000, YAMAMOTO Takashi wrote:
>>> > hi,
>>> > 
>>> > is there a near-future plan to use liblzf for any in-tree stuff?
>>> > otherwise i'd suggest to disable it because it only causes problems.
>>> > see PR/46426.
>>> 
>>> I have code that uses it, both in userspace and the kernel, but I never
>>> checked it in because I ran into a build problem with src/common.  Let me
>>> see what I can do about both issues.

any progress?

if you don't want to remove it, how about renaming it?

YAMAMOTO Takashi

>> 
>> I am looking at the PR, and I don't really think I agree that our liblzf
>> is "incompatible with the upstream version".
>> 
>> Unfortunately, the upstream version can be compiled with either of two
>> different APIs.  I think the bug is actually in programs whose autoconf
>> logic detects liblzf without detecting which API is in use.
> 
(Continue reading)

YAMAMOTO Takashi | 17 Aug 2012 02:00
Picon

Re: liblzf incompatibility

hi,

can you reply?

YAMAMOTO Takashi

> hi,
> 
>> hi,
>> 
>>> On Mon, Jun 11, 2012 at 11:45:11PM -0400, Thor Lancelot Simon wrote:
>>>> On Tue, Jun 12, 2012 at 01:58:12AM +0000, YAMAMOTO Takashi wrote:
>>>> > hi,
>>>> > 
>>>> > is there a near-future plan to use liblzf for any in-tree stuff?
>>>> > otherwise i'd suggest to disable it because it only causes problems.
>>>> > see PR/46426.
>>>> 
>>>> I have code that uses it, both in userspace and the kernel, but I never
>>>> checked it in because I ran into a build problem with src/common.  Let me
>>>> see what I can do about both issues.
> 
> any progress?
> 
> if you don't want to remove it, how about renaming it?
> 
> YAMAMOTO Takashi
> 
>>> 
>>> I am looking at the PR, and I don't really think I agree that our liblzf
(Continue reading)

Mindaugas Rasiukevicius | 28 Aug 2012 21:40
Picon

Re: liblzf incompatibility

yamt <at> mwd.biglobe.ne.jp (YAMAMOTO Takashi) wrote:
> hi,
> 
> > hi,
> > 
> >> On Mon, Jun 11, 2012 at 11:45:11PM -0400, Thor Lancelot Simon wrote:
> >>> On Tue, Jun 12, 2012 at 01:58:12AM +0000, YAMAMOTO Takashi wrote:
> >>> > hi,
> >>> > 
> >>> > is there a near-future plan to use liblzf for any in-tree stuff?
> >>> > otherwise i'd suggest to disable it because it only causes problems.
> >>> > see PR/46426.
> >>> 
> >>> I have code that uses it, both in userspace and the kernel, but I
> >>> never checked it in because I ran into a build problem with
> >>> src/common.  Let me see what I can do about both issues.
> 
> any progress?
> 
> if you don't want to remove it, how about renaming it?

I think this ought to be removed from netbsd-6 (incompatible and there are
no users, nor there is going to be as we are in RC1) and at the very least
disabled in -current.

> 
> YAMAMOTO Takashi

--

-- 
Mindaugas
(Continue reading)

Thor Lancelot Simon | 28 Aug 2012 21:52
Picon
Favicon

Re: liblzf incompatibility

On Tue, Aug 28, 2012 at 08:40:40PM +0100, Mindaugas Rasiukevicius wrote:
> 
> I think this ought to be removed from netbsd-6 (incompatible and there are
> no users, nor there is going to be as we are in RC1) and at the very least
> disabled in -current.

I would prefer to switch the userspce API to the other one and bump the
library major version number.  I do have kernel code which I am about to
check in that uses lzf (an entropy estimator).

As I told yamt privately, I will deal with this this week.  Sorry for the
delay.

Thor

Christos Zoulas | 29 Aug 2012 09:32

Re: liblzf incompatibility

In article <20120828194151.BB89014A37A <at> mail.netbsd.org>,
Mindaugas Rasiukevicius  <rmind <at> netbsd.org> wrote:
>yamt <at> mwd.biglobe.ne.jp (YAMAMOTO Takashi) wrote:
>> hi,
>> 
>> > hi,
>> > 
>> >> On Mon, Jun 11, 2012 at 11:45:11PM -0400, Thor Lancelot Simon wrote:
>> >>> On Tue, Jun 12, 2012 at 01:58:12AM +0000, YAMAMOTO Takashi wrote:
>> >>> > hi,
>> >>> > 
>> >>> > is there a near-future plan to use liblzf for any in-tree stuff?
>> >>> > otherwise i'd suggest to disable it because it only causes problems.
>> >>> > see PR/46426.
>> >>> 
>> >>> I have code that uses it, both in userspace and the kernel, but I
>> >>> never checked it in because I ran into a build problem with
>> >>> src/common.  Let me see what I can do about both issues.
>> 
>> any progress?
>> 
>> if you don't want to remove it, how about renaming it?
>
>I think this ought to be removed from netbsd-6 (incompatible and there are
>no users, nor there is going to be as we are in RC1) and at the very least
>disabled in -current.

Or just make it compatible and add _s functions that pass the state. Or
make it detect NULL state and provide the one on the stack.

(Continue reading)

Thor Lancelot Simon | 29 Aug 2012 18:30
Picon
Favicon

Re: liblzf incompatibility

On Wed, Aug 29, 2012 at 07:32:02AM +0000, Christos Zoulas wrote:
> In article <20120828194151.BB89014A37A <at> mail.netbsd.org>,
> Mindaugas Rasiukevicius  <rmind <at> netbsd.org> wrote:
> >yamt <at> mwd.biglobe.ne.jp (YAMAMOTO Takashi) wrote:
> >> hi,
> >> 
> >> > hi,
> >> > 
> >> >> On Mon, Jun 11, 2012 at 11:45:11PM -0400, Thor Lancelot Simon wrote:
> >> >>> On Tue, Jun 12, 2012 at 01:58:12AM +0000, YAMAMOTO Takashi wrote:
> >> >>> > hi,
> >> >>> > 
> >> >>> > is there a near-future plan to use liblzf for any in-tree stuff?
> >> >>> > otherwise i'd suggest to disable it because it only causes problems.
> >> >>> > see PR/46426.
> >> >>> 
> >> >>> I have code that uses it, both in userspace and the kernel, but I
> >> >>> never checked it in because I ran into a build problem with
> >> >>> src/common.  Let me see what I can do about both issues.
> >> 
> >> any progress?
> >> 
> >> if you don't want to remove it, how about renaming it?
> >
> >I think this ought to be removed from netbsd-6 (incompatible and there are
> >no users, nor there is going to be as we are in RC1) and at the very least
> >disabled in -current.
> 
> Or just make it compatible and add _s functions that pass the state. Or
> make it detect NULL state and provide the one on the stack.
(Continue reading)

Alan Barrett | 29 Aug 2012 18:59
Gravatar

Re: liblzf incompatibility

On Wed, 29 Aug 2012, Christos Zoulas wrote:
>>I think this ought to be removed from netbsd-6 (incompatible and there are
>>no users, nor there is going to be as we are in RC1) and at the very least
>>disabled in -current.
>
>Or just make it compatible and add _s functions that pass the state. Or
>make it detect NULL state and provide the one on the stack.

"_r" please, rather than "_s".  We already have several _r 
functions for "reentrant" versions of non-_r functions.  The _s 
suffix has sort of been appropriated by C201x Annex K.

--apb (Alan Barrett)

Christos Zoulas | 12 Jun 2012 05:54

Re: liblzf incompatibility

In article <20120612015812.348DC14A26D <at> mail.netbsd.org>,
YAMAMOTO Takashi <yamt <at> mwd.biglobe.ne.jp> wrote:
>hi,
>
>is there a near-future plan to use liblzf for any in-tree stuff?
>otherwise i'd suggest to disable it because it only causes problems.
>see PR/46426.
>
>YAMAMOTO Takashi

We could provide a compatible entry point instead...

christos


Gmane