Re: bug in tar 1.14-2.1

2006-03-30 Thread Bdale Garbee
On Tue, 2006-03-28 at 05:27 +0200, Goswin von Brederlow wrote:

 Again? I wrote a bug about this years ago with a fix. I think is was
 just adding --rsh=/usr/bin/rsh to the configure call.

Your patch in response to 185594 added an RSH environment variable
definition to the configure invocation.  The problem now is that if the
'make' invocation ends up re-running configure, that definition isn't
present.  So, the easy fix is to define the same environment variable
for the make invocation.

Bdale


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-27 Thread Andreas Barth
* Martin Zobel-Helas ([EMAIL PROTECTED]) [060324 16:00]:
 Looks like just rebuilding the security version resolves that error, for
 whatever reason. Julien and me just cross checked that and got the same
 result.
 
 If noone minds we reupload tar with a bumped version number to s-p-u.

Is a binary-only upload enough? If so, why not just queue a binNMU by
the buildd? (And one should check all the archs BTW, and also add a test
suite one day :)


Cheers,
Andi
-- 
  http://home.arcor.de/andreas-barth/


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-27 Thread Martin Zobel-Helas
Hi Andi,

On Monday, 27 Mar 2006, you wrote:
 * Martin Zobel-Helas ([EMAIL PROTECTED]) [060324 16:00]:
  Looks like just rebuilding the security version resolves that error, for
  whatever reason. Julien and me just cross checked that and got the same
  result.
  
  If noone minds we reupload tar with a bumped version number to s-p-u.
 
 Is a binary-only upload enough? If so, why not just queue a binNMU by
 the buildd? (And one should check all the archs BTW, and also add a test
 suite one day :)

as Julien and me found out, tar works only if either ssh is installed or
the correct enviroment variables are set. As ssh is not installed per
default in buildd enviroment we need to patch the rules-file to get the
correct enviroment variables set.

So, no, binNMU is not enough (only if you can persue all buildd
maintainers to install ssh inside the changeroot per default ;) )

Greetings
Martin


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-27 Thread Goswin von Brederlow
Martin Zobel-Helas [EMAIL PROTECTED] writes:

 Hi Andi,

 On Monday, 27 Mar 2006, you wrote:
 * Martin Zobel-Helas ([EMAIL PROTECTED]) [060324 16:00]:
  Looks like just rebuilding the security version resolves that error, for
  whatever reason. Julien and me just cross checked that and got the same
  result.
  
  If noone minds we reupload tar with a bumped version number to s-p-u.
 
 Is a binary-only upload enough? If so, why not just queue a binNMU by
 the buildd? (And one should check all the archs BTW, and also add a test
 suite one day :)

 as Julien and me found out, tar works only if either ssh is installed or
 the correct enviroment variables are set. As ssh is not installed per
 default in buildd enviroment we need to patch the rules-file to get the
 correct enviroment variables set.

 So, no, binNMU is not enough (only if you can persue all buildd
 maintainers to install ssh inside the changeroot per default ;) )

 Greetings
 Martin

Again? I wrote a bug about this years ago with a fix. I think is was
just adding --rsh=/usr/bin/rsh to the configure call.

MfG
Goswin


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-25 Thread Bdale Garbee
On Fri, 2006-03-24 at 22:07 +0100, Julien Danjou wrote:

 Finally, I solved it this way:
 
 diff -u tar-1.14/debian/rules tar-1.14/debian/rules
 --- tar-1.14/debian/rules
 +++ tar-1.14/debian/rules
 @@ -16,7 +16,7 @@
  build-stamp:
 dh_testdir
  
 -   $(MAKE)
 +   RSH=/usr/bin/rsh CFLAGS=-O2 -g -Wall $(MAKE)
 (cd tests ; $(MAKE) check-TESTS)
  
 touch build-stamp

It certainly seems to make sense to provide the same environment to make
that we're using for configure, so I'll add this to my CVS for the next
upload to unstable.  If that is sufficient to ensure a clean build of a
1.14-2.2 version on all architectures affected by the bug in question, I
have no problem with that change being included for the next stable
update.

However, I don't immediately understand why the auto* tools are being
invoked by make during this build?  Reviewing the tar changelog, I made
explicit reference for 1.14-1 that autoconf and automake build
dependencies were elminated.  Nothing I changed for 1.14-2 should have
affected that, and it's not obvious that anything in the diff between -2
and -2.1 should have affected that either.  However, I left in place the
code in the debian/rules file that updates config.sub and config.guess
in the 'clean' target *if* autotools-dev is installed.  So, I wonder if
the underlying problem is that those files are being updated and somehow
triggering this behavior?

Goswin indicates in his reply to this message that we should touch the
files in the right order, but I'm not sure what that refers to in this
case?

For 1.15.1-4, I re-introduced a dependency on autoconf in response to
#354194.  In hindsight, that may not have been the right thing to do...
and even if it was the right thing to do, I'm not sure it was really
sufficient.  It would seem desireable to not run the auto* tools on
every build if we don't need to, yet if we're going to then there should
probably be an autotools-dev build dependency, etc?

For whatever reason, I'm having a hard time thinking clearly about this
today, and so advice and/or suggested patches for the version currently
in unstable to do the right thing in the future would be appreciated.

Regards,

Bdale


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-25 Thread Andreas Metzler
On 2006-03-25 Bdale Garbee [EMAIL PROTECTED] wrote:
[...]
 However, I don't immediately understand why the auto* tools are being
 invoked by make during this build?
[...]

[EMAIL PROTECTED]:/tmp$ lsdiff -z tar_1.14-2.1.diff.gz | \
 egrep -v '/debian/|/src/'
tar-1.14/config/config.guess
tar-1.14/config/config.sub
tar-1.14/po/de.po
tar-1.14/configure.ac
tar-1.14/doc/tar.texi

You are patching configure.ac, which causes auto* to try to regenerate
./configure. Remove the patchlet and it should build fine. As you are not
shipping the corresponding change to ./configure the thing is
ineffictive anyway, unless you are rebuildung on a system that has the
whole auto* shebang installed.[1]

hth, cu andreas

[1] It actually FTBFS for me, with auto* installed:
cd .  /bin/bash /tmp/tar-1.14/config/missing --run aclocal-1.8 -I m4
aclocal: m4/gettext.m4: 378: macro `AM_LC_MESSAGES' not found in library
aclocal: macro `jm_GLIBC21' required but not defined
make: *** [aclocal.m4] Error 1
-- 
The 'Galactic Cleaning' policy undertaken by Emperor Zhark is a personal
vision of the emperor's, and its inclusion in this work does not constitute
tacit approval by the author or the publisher for any such projects,
howsoever undertaken.(c) Jasper Ffforde


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-25 Thread Goswin von Brederlow
Andreas Metzler [EMAIL PROTECTED] writes:

 On 2006-03-25 Bdale Garbee [EMAIL PROTECTED] wrote:
 [...]
 However, I don't immediately understand why the auto* tools are being
 invoked by make during this build?
 [...]

 [EMAIL PROTECTED]:/tmp$ lsdiff -z tar_1.14-2.1.diff.gz | \
  egrep -v '/debian/|/src/'
 tar-1.14/config/config.guess
 tar-1.14/config/config.sub
 tar-1.14/po/de.po
 tar-1.14/configure.ac
 tar-1.14/doc/tar.texi

 You are patching configure.ac, which causes auto* to try to regenerate
 ./configure. Remove the patchlet and it should build fine. As you are not
 shipping the corresponding change to ./configure the thing is
 ineffictive anyway, unless you are rebuildung on a system that has the
 whole auto* shebang installed.[1]

 hth, cu andreas

 [1] It actually FTBFS for me, with auto* installed:
 cd .  /bin/bash /tmp/tar-1.14/config/missing --run aclocal-1.8 -I m4
 aclocal: m4/gettext.m4: 378: macro `AM_LC_MESSAGES' not found in library
 aclocal: macro `jm_GLIBC21' required but not defined
 make: *** [aclocal.m4] Error 1

And remeber, if you ship a patch for configure and configure.ac then
the files will have timestamps in the same order as they are in the
diff, which is probably random.

To make sure the timestamps fit you have to touch configure.ac; touch
configure in rules then.

MfG
Goswin


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-25 Thread Bdale Garbee
On Sat, 2006-03-25 at 18:36 +0100, Andreas Metzler wrote:
 On 2006-03-25 Bdale Garbee [EMAIL PROTECTED] wrote:
 [...]
  However, I don't immediately understand why the auto* tools are being
  invoked by make during this build?
 [...]
 
 [EMAIL PROTECTED]:/tmp$ lsdiff -z tar_1.14-2.1.diff.gz | \
  egrep -v '/debian/|/src/'
 tar-1.14/config/config.guess
 tar-1.14/config/config.sub
 tar-1.14/po/de.po
 tar-1.14/configure.ac
 tar-1.14/doc/tar.texi
 
 You are patching configure.ac, which causes auto* to try to regenerate
 ./configure.  Remove the patchlet and it should build fine. 

Aha!  Thanks.  Somehow, I missed that.  The only change was part of a
patch from upstream to replace a buggy use of valloc with use of malloc
instead, such that configure no longer needed to test for the presence
of valloc.  I can see no problem with undoing that change in
configure.ac and touching configure.

 [1] It actually FTBFS for me, with auto* installed:

Interesting.  I'm not very motivated to chase that down in 1.14, if for
no other reason than unstable is using 1.15 at this point. 

Bdale


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-24 Thread Martin Zobel-Helas
Hi mollo,

On Sunday, 19 Mar 2006, you wrote:
 On Tue, 7 Mar 2006 15:19:58 +0100

 using tar 1.14-2.1
 
 fw:/home/mathieu# tar --rmt-command=/usr/sbin/rmt -cvf
 '[EMAIL PROTECTED]:/home/mathieu/test.tgz' /etc tar:
 [EMAIL PROTECTED]:/home/mathieu/test.tgz: Cannot open: Input/output
 error tar: Error is not recoverable: exiting now

ack, same here. i can reproduce that error.

 
 
 using old tar 1.14-2 :
 
 fw:/home/mathieu# tar.ori --rmt-command=/usr/sbin/rmt -cvf
 '[EMAIL PROTECTED]:/home/mathieu/test.tgz' /etc/ssh Password: 
 tar.ori: Removing leading `/' from member names
 /etc/ssh/

Looks like just rebuilding the security version resolves that error, for
whatever reason. Julien and me just cross checked that and got the same
result.

If noone minds we reupload tar with a bumped version number to s-p-u.

Greetings
Martin


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-24 Thread Bdale Garbee
On Fri, 2006-03-24 at 15:53 +0100, Martin Zobel-Helas wrote:

 If noone minds we reupload tar with a bumped version number to s-p-u.

Ok with me.

Bdale


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Re: bug in tar 1.14-2.1

2006-03-24 Thread Julien Danjou
On Fri, Mar 24, 2006 at 03:53:03PM +0100, Martin Zobel-Helas wrote:
 Looks like just rebuilding the security version resolves that error, for
 whatever reason. Julien and me just cross checked that and got the same
 result.

We tried to reproduce the bug with zobel, and finally discovered the Truth.

I'll try to explain:

$ apt-get source tar=1.14-2.1
$ cd tar-1.14
$ debuild  ../first-build.log
$ grep REMOTE_SHELL config.h
/* #undef REMOTE_SHELL */
$ debuild  ../second-build.log
$ grep REMOTE_SHELL config.h
#define REMOTE_SHELL /usr/bin/rsh

And with the diff between log files it is obvious, in the first run we
can see:

/usr/bin/make
make[1]: Entering directory `/home/tar-1.14'
cd .  /bin/sh /home/tar-1.14/config/missing --run aclocal-1.8 -I m4
/home/tar-1.14/config/missing: line 52: aclocal-1.8: command not found
WARNING: `aclocal-1.8' is missing on your system.  You should only need it if
 you modified `acinclude.m4' or `configure.ac'.  You might want
 to install the `Automake' and `Perl' packages.  Grab them from
 any GNU archive site.
 cd .  /bin/sh /home/tar-1.14/config/missing --run automake-1.8 --gnits 
/home/tar-1.14/config/missing: line 52: automake-1.8: command not found
WARNING: `automake-1.8' is missing on your system.  You should only need it if
 you modified `Makefile.am', `acinclude.m4' or `configure.ac'.
 You might want to install the `Automake' and `Perl' packages.
 Grab them from any GNU archive site.
cd .  /bin/sh /home/tar-1.14/config/missing --run autoconf
/home/tar-1.14/config/missing: line 52: autoconf: command not found
WARNING: `autoconf' is missing on your system.  You should only need it if
 you modified `configure.ac'.  You might want to install the
 `Autoconf' and `GNU m4' packages.  Grab them from any GNU
 archive site.
/bin/sh ./config.status --recheck
[...]

So ./configure is launched again without $RSH environment variable set
correctly.

Finally, I solved it this way:

diff -u tar-1.14/debian/rules tar-1.14/debian/rules
--- tar-1.14/debian/rules
+++ tar-1.14/debian/rules
@@ -16,7 +16,7 @@
 build-stamp:
dh_testdir
 
-   $(MAKE)
+   RSH=/usr/bin/rsh CFLAGS=-O2 -g -Wall $(MAKE)
(cd tests ; $(MAKE) check-TESTS)
 
touch build-stamp


Happy birthday Martin ! ;-)

Regards,
-- 
Julien Danjou
.''`.  Debian Developer
: :' : http://julien.danjou.info
`. `'  http://people.debian.org/~acid
  `-   9A0D 5FD9 EB42 22F6 8974  C95C A462 B51E C2FE E5CD


signature.asc
Description: Digital signature


Re: bug in tar 1.14-2.1

2006-03-24 Thread Goswin von Brederlow
Julien Danjou [EMAIL PROTECTED] writes:

 On Fri, Mar 24, 2006 at 03:53:03PM +0100, Martin Zobel-Helas wrote:
 Looks like just rebuilding the security version resolves that error, for
 whatever reason. Julien and me just cross checked that and got the same
 result.

 We tried to reproduce the bug with zobel, and finally discovered the Truth.

 I'll try to explain:

 $ apt-get source tar=1.14-2.1
 $ cd tar-1.14
 $ debuild  ../first-build.log
 $ grep REMOTE_SHELL config.h
 /* #undef REMOTE_SHELL */
 $ debuild  ../second-build.log
 $ grep REMOTE_SHELL config.h
 #define REMOTE_SHELL /usr/bin/rsh

 And with the diff between log files it is obvious, in the first run we
 can see:

 /usr/bin/make
 make[1]: Entering directory `/home/tar-1.14'
 cd .  /bin/sh /home/tar-1.14/config/missing --run aclocal-1.8 -I m4
 /home/tar-1.14/config/missing: line 52: aclocal-1.8: command not found
 WARNING: `aclocal-1.8' is missing on your system.  You should only need it if
  you modified `acinclude.m4' or `configure.ac'.  You might want
  to install the `Automake' and `Perl' packages.  Grab them from
  any GNU archive site.
  cd .  /bin/sh /home/tar-1.14/config/missing --run automake-1.8 --gnits 
 /home/tar-1.14/config/missing: line 52: automake-1.8: command not found
 WARNING: `automake-1.8' is missing on your system.  You should only need it if
  you modified `Makefile.am', `acinclude.m4' or `configure.ac'.
  You might want to install the `Automake' and `Perl' packages.
  Grab them from any GNU archive site.
 cd .  /bin/sh /home/tar-1.14/config/missing --run autoconf
 /home/tar-1.14/config/missing: line 52: autoconf: command not found
 WARNING: `autoconf' is missing on your system.  You should only need it if
  you modified `configure.ac'.  You might want to install the
  `Autoconf' and `GNU m4' packages.  Grab them from any GNU
  archive site.
 /bin/sh ./config.status --recheck
 [...]

You should touch the files in the right order so the timestamps match.

MfG
Goswin


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]