Debian Bug report logs - #880166
regression in dpkg-gensymbols output on symbols changes

Reply or subscribe to this bug.

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Rik Mills <rikmills@kubuntu.org>

To: submit@bugs.debian.org

Subject: regression in dpkg-gensymbols output on symbols changes

Date: Mon, 30 Oct 2017 09:03:06 +0000

Package: dpkg
Version: 1.19.0.4

The output of dpkg-gensymbols has regressed when reporting new or
changed symbols

Previously the output stated:

dpkg-gensymbols: debian/libnameDEBIAN/symbols doesn't match completely
debian/libname.symbols

this now incorrectly states:

dpkg-gensymbols: warning: no debian/symbols file used as basis for
generating debian/libname/DEBIAN/symbols

Message #10 received at 880166@bugs.debian.org (full text, mbox, reply):

From: Dmitry Shachnev <mitya57@debian.org>

To: Rik Mills <rikmills@kubuntu.org>, 880166@bugs.debian.org

Subject: Re: Bug#880166: regression in dpkg-gensymbols output on symbols changes

Date: Mon, 30 Oct 2017 12:13:12 +0300

[Message part 1 (text/plain, inline)]

On Mon, Oct 30, 2017 at 09:03:06AM +0000, Rik Mills wrote:
> The output of dpkg-gensymbols has regressed when reporting new or
> changed symbols
>
> Previously the output stated:
>
> dpkg-gensymbols: debian/libnameDEBIAN/symbols doesn't match completely
> debian/libname.symbols
>
> this now incorrectly states:
>
> dpkg-gensymbols: warning: no debian/symbols file used as basis for
> generating debian/libname/DEBIAN/symbols

The attached small patch fixes the issue for me. Tested with libjpeg-turbo,
where one symbol was deliberately removed.

The previous if (defined $state) check was always true, because a //=
operator is used to define $state two lines above.

--
Dmitry Shachnev

[880166.diff (text/x-diff, attachment)]

[signature.asc (application/pgp-signature, inline)]

Message #15 received at 880166@bugs.debian.org (full text, mbox, reply):

From: Rik Mills <rikmills@kubuntu.org>

To: Dmitry Shachnev <mitya57@debian.org>, 880166@bugs.debian.org

Subject: Re: Bug#880166: regression in dpkg-gensymbols output on symbols changes

Date: Mon, 30 Oct 2017 10:51:48 +0000

On 30/10/17 09:13, Dmitry Shachnev wrote:
> The attached small patch fixes the issue for me. Tested with libjpeg-turbo,
> where one symbol was deliberately removed.
> 
> The previous if (defined $state) check was always true, because a //=
> operator is used to define $state two lines above.
> 
> --
> Dmitry Shachnev
> 

This patch appears to fix the issue for me on the libjpeg-turbo test
case, and builds on the kubuntu CI which was having issues on not seeing
the correct output in build logs.

Message #20 received at 880166@bugs.debian.org (full text, mbox, reply):

From: Rik Mills <rikmills@kubuntu.org>

To: 880166@bugs.debian.org

Subject: Re: Bug#880166: regression in dpkg-gensymbols output on symbols changes

Date: Tue, 31 Oct 2017 09:01:59 +0000

Also tracked in:
https://bugs.launchpad.net/ubuntu/+source/dpkg/+bug/1728847

Message #25 received at 880166@bugs.debian.org (full text, mbox, reply):

From: Rik Mills <rikmills@kubuntu.org>

To: 880166@bugs.debian.org

Subject: Re: Bug#880166: regression in dpkg-gensymbols output on symbols changes

Date: Mon, 1 Jan 2018 10:57:16 +0000

Hi, and happy new year

Could the patch from Dmitry Shachnev applied to dpkg soonish if that is ok?

I have patched it for Kubuntu's CI repositories, so temporarily our
symbols handling is not messed up, but it would obviously be better if
the correct output could be fixed at debian's end and then synced to the
main ubuntu archive.

Thank you

Rik Mills
Kubuntu Developer

Send a report that this bug log contains spam.