[HEADS UP] ar(1) front-end committed. (notes for cross compile)

classic Classic list List threaded Threaded
61 messages Options
1234
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Kai Wang
On Sun, Feb 24, 2008 at 10:10:04AM -0800, David O'Brien wrote:
> On Fri, Feb 22, 2008 at 06:00:07PM +0100, Kai Wang wrote:
> > Would it be better if we call them gar and granlib?  Solaris did
> > that. Also if I remember correctly, some ports probes gar.  We also
> > call GNU make as gmake...
>
> Why do we want
>
> I don't like "gar" as that is pronounceable to the point I could easily
> see that being the real name of an existing program.

Yes, you've got a good point. Actually I'm fine with either way...

> Also, why do we want ports using gnu-ar specifically vs. what ever is our
> native 'ar'?  If our native 'ar' isn't up to the task, we shouldn't be
> doing this endeavor at all.
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Ruslan Ermilov
In reply to this post by David O'Brien
On Sun, Feb 24, 2008 at 10:04:33AM -0800, David O'Brien wrote:
> On Sat, Feb 23, 2008 at 11:18:08PM +0300, Ruslan Ermilov wrote:
> > now bootstrap BSD ar on systems >700044, and that we call
> > GNU ar/ranlib with the "g" prefix instead of "gnu-".
>
> Why are you going against my preferences for "gnu-" - if I liked "g"
> I would have done it that way in my patch.
>
The reasoning for the "g" prefix sounded sane to me.
Anyway, I don't want to argue about this point, and
will leave it to you to decide (I've reverted to the
"gnu-" prefix).

> Its seems those that have expressed an opinion want to switch to the new
> 'ar' ASAP.  So why not this patch?
>
Because your patch won't allow to cross-build on
systems predating 700044.

> (BTW, what is the sort order in Makefile.inc1?  BOOTSTRAPPING date and
> alphabetical?)
>
The intended order was by a pathname.  It's missorted
now.


Cheers,
--
Ruslan Ermilov
[hidden email]
FreeBSD committer
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Dominic Fandrey-2
Ruslan Ermilov wrote:

> On Sun, Feb 24, 2008 at 10:04:33AM -0800, David O'Brien wrote:
>> On Sat, Feb 23, 2008 at 11:18:08PM +0300, Ruslan Ermilov wrote:
>>> now bootstrap BSD ar on systems >700044, and that we call
>>> GNU ar/ranlib with the "g" prefix instead of "gnu-".
>> Why are you going against my preferences for "gnu-" - if I liked "g"
>> I would have done it that way in my patch.
>>
> The reasoning for the "g" prefix sounded sane to me.
> Anyway, I don't want to argue about this point, and
> will leave it to you to decide (I've reverted to the
> "gnu-" prefix).

I'd prefer the g-prefix. It's commonly used, just thing gmake, gawk and
similar stuff.
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

David O'Brien
In reply to this post by Ruslan Ermilov
On Sun, Feb 24, 2008 at 10:06:37PM +0300, Ruslan Ermilov wrote:
> On Sun, Feb 24, 2008 at 10:04:33AM -0800, David O'Brien wrote:
> > Its seems those that have expressed an opinion want to switch to the new
> > 'ar' ASAP.  So why not this patch?
>
> Because your patch won't allow to cross-build on
> systems predating 700044.

Why not?  If ${BOOTSTRAPPING} >= 700044 we add usr.bin/ar to the list.

What is your patch to just get rid of the WITH_GNUAR/WITH_BSDAR similar
to mine?

--
-- David  ([hidden email])
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Ruslan Ermilov
On Sun, Feb 24, 2008 at 11:32:21AM -0800, David O'Brien wrote:

> On Sun, Feb 24, 2008 at 10:06:37PM +0300, Ruslan Ermilov wrote:
> > On Sun, Feb 24, 2008 at 10:04:33AM -0800, David O'Brien wrote:
> > > Its seems those that have expressed an opinion want to switch to the new
> > > 'ar' ASAP.  So why not this patch?
> >
> > Because your patch won't allow to cross-build on
> > systems predating 700044.
>
> Why not?  If ${BOOTSTRAPPING} >= 700044 we add usr.bin/ar to the list.
>
I said predating 700044.

> What is your patch to just get rid of the WITH_GNUAR/WITH_BSDAR similar
> to mine?
>
ENOPARSE.

How's the attached patch?  The commit log would be:

: Make again BSD ar(1) the default system ar(1), now properly handling
: source upgrades by falling back to GNU ar(1) as necessary.  Option
: WITH_BSDAR is gone.  Option _WITH_GNUAR to aid in upgrades is NOT
: supposed to be set by the user.  Stop bootstrapping BSD ar(1) on the
: next __FreeBSD_version bump, as there are no known bugs in it.  Bump
: __FreeBSD_version to anticipate this and to flag the switch to BSD
: ar(1), should it be needed for something.
:
: Input from: obrien, des, kaiw


Cheers,
--
Ruslan Ermilov
[hidden email]
FreeBSD committer

_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

David O'Brien
On Sun, Feb 24, 2008 at 10:43:58PM +0300, Ruslan Ermilov wrote:

> On Sun, Feb 24, 2008 at 11:32:21AM -0800, David O'Brien wrote:
> > On Sun, Feb 24, 2008 at 10:06:37PM +0300, Ruslan Ermilov wrote:
> > > On Sun, Feb 24, 2008 at 10:04:33AM -0800, David O'Brien wrote:
> > > > Its seems those that have expressed an opinion want to switch to the new
> > > > 'ar' ASAP.  So why not this patch?
> > >
> > > Because your patch won't allow to cross-build on
> > > systems predating 700044.
> >
> > Why not?  If ${BOOTSTRAPPING} >= 700044 we add usr.bin/ar to the list.
> >
> I said predating 700044.

*sigh* what's 700044 - it would be nice if you'd say what that date
represents so one doens't have to dig around www.freebsd.org to figure it
out.


> > What is your patch to just get rid of the WITH_GNUAR/WITH_BSDAR similar
> > to mine?
> >
> ENOPARSE.
>
> How's the attached patch?  The commit log would be:

I don't think you answered me - why do we need the WITH_* stuff at all.
Just install GNU ar as gnu-ar from now on and have the build system
handle it.

_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

David O'Brien
In reply to this post by Ruslan Ermilov
On Sun, Feb 24, 2008 at 10:43:58PM +0300, Ruslan Ermilov wrote:

> On Sun, Feb 24, 2008 at 11:32:21AM -0800, David O'Brien wrote:
> > On Sun, Feb 24, 2008 at 10:06:37PM +0300, Ruslan Ermilov wrote:
> > > On Sun, Feb 24, 2008 at 10:04:33AM -0800, David O'Brien wrote:
> > > > Its seems those that have expressed an opinion want to switch to the new
> > > > 'ar' ASAP.  So why not this patch?
> > >
> > > Because your patch won't allow to cross-build on
> > > systems predating 700044.
> >
> > Why not?  If ${BOOTSTRAPPING} >= 700044 we add usr.bin/ar to the list.
> >
> I said predating 700044.

What is the thing you're checking the value 700044 for?

    http://www.freebsd.org/doc/en_US.ISO8859-1/books/porters-handbook/freebsd-versions.html
    700044
        7.0-CURRENT after changing the argument for vn_open()/VOP_OPEN()
        from filedescriptor index to the struct file *.
    700045
        7.0-CURRENT after changing pam_nologin(8) to provide an account
        management function instead of an authentication function to the
        PAM framework.

_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Kai Wang
On Sun, Feb 24, 2008 at 12:51:21PM -0800, David O'Brien wrote:

> What is the thing you're checking the value 700044 for?
>
>     http://www.freebsd.org/doc/en_US.ISO8859-1/books/porters-handbook/freebsd-versions.html
>     700044
>         7.0-CURRENT after changing the argument for vn_open()/VOP_OPEN()
>         from filedescriptor index to the struct file *.
>     700045
>         7.0-CURRENT after changing pam_nologin(8) to provide an account
>         management function instead of an authentication function to the
>         PAM framework.
>

I guess, probably des@ meant that: libarchive 2.2.3 (committed
at May 29 2007) included the last major fix to the 'ar' support,
and 700044 (though not specific to this) bumped at Jun 7 2007
is closest to that day?

Kai

_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

David O'Brien
On Sun, Feb 24, 2008 at 10:26:42PM +0100, Kai Wang wrote:

> On Sun, Feb 24, 2008 at 12:51:21PM -0800, David O'Brien wrote:
> > What is the thing you're checking the value 700044 for?
> >     http://www.freebsd.org/doc/en_US.ISO8859-1/books/porters-handbook/freebsd-versions.html
> >     700044
> >         7.0-CURRENT after changing the argument for vn_open()/VOP_OPEN()
> >         from filedescriptor index to the struct file *.
> >     700045
> >         7.0-CURRENT after changing pam_nologin(8) to provide an account
> >         management function instead of an authentication function to the
> >         PAM framework.
>
> I guess, probably des@ meant that: libarchive 2.2.3 (committed
> at May 29 2007) included the last major fix to the 'ar' support,
> and 700044 (though not specific to this) bumped at Jun 7 2007
> is closest to that day?

If that is the case - this web page should be updated with that
information.  It is find to go back after the fact and note that a
particular version number represents more than one thing.

--
-- David  ([hidden email])
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Dag-Erling Smørgrav
In reply to this post by David O'Brien
"David O'Brien" <[hidden email]> writes:
> I don't like "gar" as that is pronounceable to the point I could easily
> see that being the real name of an existing program.

I've told you already that there is no such program in ports.  In
addition, no project by that name is registered in SourceForge, cia.vc
or ohloh.net.  I don't see the point of being inconsistent just because
you can pronounce "gar".

(BTW, I can pronounce "groff".  Should we rename it to gnu-roff?)

DES
--
Dag-Erling Smørgrav - [hidden email]
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Dag-Erling Smørgrav
Dag-Erling Smørgrav <[hidden email]> writes:
> (BTW, I can pronounce "groff".  Should we rename it to gnu-roff?)

not to mention gawk...

DES
--
Dag-Erling Smørgrav - [hidden email]
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"

Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Ruslan Ermilov
In reply to this post by David O'Brien
On Sun, Feb 24, 2008 at 01:34:34PM -0800, David O'Brien wrote:

> On Sun, Feb 24, 2008 at 10:26:42PM +0100, Kai Wang wrote:
> > On Sun, Feb 24, 2008 at 12:51:21PM -0800, David O'Brien wrote:
> > > What is the thing you're checking the value 700044 for?
> > >     http://www.freebsd.org/doc/en_US.ISO8859-1/books/porters-handbook/freebsd-versions.html
> > >     700044
> > >         7.0-CURRENT after changing the argument for vn_open()/VOP_OPEN()
> > >         from filedescriptor index to the struct file *.
> > >     700045
> > >         7.0-CURRENT after changing pam_nologin(8) to provide an account
> > >         management function instead of an authentication function to the
> > >         PAM framework.
> >
> > I guess, probably des@ meant that: libarchive 2.2.3 (committed
> > at May 29 2007) included the last major fix to the 'ar' support,
> > and 700044 (though not specific to this) bumped at Jun 7 2007
> > is closest to that day?
>
> If that is the case - this web page should be updated with that
> information.  It is find to go back after the fact and note that a
> particular version number represents more than one thing.
>
We've been traditionally (ab)using __FreeBSD_version to bootstrap
things in this way, without necessarily bumping __FreeBSD_version.
There are two reasons: 1) we often don't know if a particular bug
or change affects bootstrapping, so we don't always bump it.  For
example, a change in libc may cause some utility to be put to the
list of bootstrap-tools, and we don't discover it until the
version is bumped for another reason.  2) there's no harm in
bootstrapping more than necessary.


Cheers,
--
Ruslan Ermilov
[hidden email]
FreeBSD committer
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

David O'Brien
In reply to this post by Dag-Erling Smørgrav
On Mon, Feb 25, 2008 at 03:41:45AM +0100, Dag-Erling Smrgrav wrote:
> "David O'Brien" <[hidden email]> writes:
> > I don't like "gar" as that is pronounceable to the point I could easily
> > see that being the real name of an existing program.
>
> I've told you already that there is no such program in ports.  In
> addition, no project by that name is registered in SourceForge, cia.vc
> or ohloh.net.

Did you look in every ~/bin and /usr/local/bin?

> I don't see the point of being inconsistent just because
> you can pronounce "gar".

DES - why do you care??  Are you planning on invoking it directly?

Additionally from what's been said, various autoconf configs may even
explicity look for gar and use it instead of /usr/bin/ar.  Given your
desire to replace GNU ar with this new ar, why would you not want to use
it everywhere?

--
-- David  ([hidden email])
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

David O'Brien
In reply to this post by Ruslan Ermilov
On Mon, Feb 25, 2008 at 10:18:25AM +0300, Ruslan Ermilov wrote:
> We've been traditionally (ab)using __FreeBSD_version to bootstrap
> things in this way, without necessarily bumping __FreeBSD_version.
> There are two reasons: 1) we often don't know if a particular bug
> or change affects bootstrapping, so we don't always bump it.  For
> example, a change in libc may cause some utility to be put to the
> list of bootstrap-tools, and we don't discover it until the
> version is bumped for another reason.

Right, but once you've attached added significance to a particular
__FreeBSD_version (after the fact) it really should be documented
in the Porter's Handbook.  At least I used to go add to a particular
__FreeBSD_version entry once I started using that value for something
other than the original bumping reason.


> 2) there's no harm in bootstrapping more than necessary.

Duplicated effort, and long build world times.

--
-- David  ([hidden email])
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Astrodog
In as far as gnu-ar vs gar goes... I want mine blue, thanks.

I think being able to decide which "ar" one uses could be fairly
important. I'm not entirely confident that now, and forever more, they
will have identical functionality.

--- Harrison
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Dag-Erling Smørgrav
In reply to this post by David O'Brien
"David O'Brien" <[hidden email]> writes:
> Ruslan Ermilov <[hidden email]> writes:
> > 2) there's no harm in bootstrapping more than necessary.
> Duplicated effort, and long build world times.

Uh, David, it's only a few days, and it only adds a couple of minutes at
most for those upgrading in the window between the libarchive fix and
7000044.  You're splitting hairs - although I agree about documenting it.

DES
--
Dag-Erling Smørgrav - [hidden email]
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Dag-Erling Smørgrav
In reply to this post by David O'Brien
"David O'Brien" <[hidden email]> writes:
> Additionally from what's been said, various autoconf configs may even
> explicity look for gar and use it instead of /usr/bin/ar.  Given your
> desire to replace GNU ar with this new ar, why would you not want to use
> it everywhere?

1) I doubt it 2) bsd.ports.mk already addresses this for gmake by
explicitly setting MAKE=gmake, it is trivial to extend this with
AR=/usr/bin/ar and RANLIB=/usr/bin/ranlib if you're so worried.

This is getting really counter-productive...

DES
--
Dag-Erling Smørgrav - [hidden email]
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

Ruslan Ermilov
In reply to this post by David O'Brien
On Mon, Feb 25, 2008 at 12:35:55AM -0800, David O'Brien wrote:
> > 2) there's no harm in bootstrapping more than necessary.
>
> Duplicated effort, and long build world times.
>
Ahem.  It's less than two seconds in wall clock time on a
modern hardware.  :-)


Cheers,
--
Ruslan Ermilov
[hidden email]
FreeBSD committer
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

David O'Brien
On Mon, Feb 25, 2008 at 02:39:51PM +0300, Ruslan Ermilov wrote:
> On Mon, Feb 25, 2008 at 12:35:55AM -0800, David O'Brien wrote:
> > > 2) there's no harm in bootstrapping more than necessary.
> >
> > Duplicated effort, and long build world times.
>
> Ahem.  It's less than two seconds in wall clock time on a
> modern hardware.  :-)

Unless we build everything as a bootstrap tool (related to Marcel's
comment in this thread).

Do you have a new diff that doesn't use any of the WITH_* stuff and just
converts to using the new 'ar'?
 
--
-- David  ([hidden email])
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
Reply | Threaded
Open this post in threaded view
|

Re: [HEADS UP] ar(1) front-end committed. (notes for cross compile)

David O'Brien
In reply to this post by Dag-Erling Smørgrav
On Mon, Feb 25, 2008 at 11:16:32AM +0100, Dag-Erling Smrgrav wrote:
> "David O'Brien" <[hidden email]> writes:
> > Ruslan Ermilov <[hidden email]> writes:
> > > 2) there's no harm in bootstrapping more than necessary.
> > Duplicated effort, and long build world times.
>
> Uh, David, it's only a few days, and it only adds a couple of minutes at
> most for those upgrading in the window between the libarchive fix and
> 7000044.  You're splitting hairs - although I agree about documenting it.

No - you're misunderstanding me.  ALL I'm saying is when one uses a
particular __FreeBSD_version they should document the significance of
that version in the Porter's Handbook.  One should be able to look up
why that particular number was used.

I am not in the lest bit complaining that the value 7000044 isn't the
exact second the event ru is testing for occurred.

--
-- David  ([hidden email])
_______________________________________________
[hidden email] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "[hidden email]"
1234