Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
On 19.06.2013 22:42, Dmitry Smirnov wrote: Hi Daniel, Thanks for your recent corrections. Just few more left to do and I'll upload for you. Thanks for additional pointers and the close review - the package is going to be much better than it's Ubuntu precursor. *** control *** * When I mentioned unnecessary versioned dependencies I meant that versioning is unnecessary and safe to drop. There is no point writing cmake (= 2.8.2+dfsg.1-0+squeeze1) when just cmake is enough because cmake in oldstable is above minimum requirement. Same applies to libqt4-dev. * Short package description should start with lowercase letter. Fixed. *** copyright *** 1) * Upstream uses two emails in source files: Joseph Artsimovich joseph.artsimov...@gmail.com Joseph Artsimovich josep...@mail.ru The latter statement appears to be older (it is often accompanied by years 2007-2008 and one would hardly use this horrible russian email service while having account at gmail). I would at least mention first statement (i.e. contact email) in Upstream-Contact or even include both statement to copyright paragraph as it would be speculative to decide which statement is more important/valid. I've been in contact with Joseph over joseph.artsimov...@gmail.com a couple of weeks ago. I've just changed that throughout, also in the manpages. 2) * The following copyright holders are not listed in debian/copyright: Vadim Kuznetsov ()DikBSD dik...@gmail.com 2011 Petr Kovar pej...@gmail.com I've added individual paragraphs for these files. 3) * Based on code from the GIMP project is not a copyright statement so perhaps it's better to move it to Comment section. Then it would become obvious that word Copyright is used twice in the paragraph. Yes, I've moved the based on info of this files to Comments. 4) * At least one file in packaging/osx is under BOOST license. Also file ScanTailor.icns is a (sourceless?) binary file... I've added a licence paragraph for packaging/osx/user-config.jam 5) * License: PD perhaps is better written as License: public-domain. For files in Public-Domain copyright is not applicable so it is better to write Copyright: not-applicable but mention origin of files (i.e. Tango project and URL) in Comment field. Changed that. 6) * One of the man pages is written by Artem Popov art...@gmail.com but that's not how his name is written in debian/copyright. It is always safer to use original statement as written by author without unnecessary translation of person's name to native characters even if it is done correctly. Yes, all right, changed that. *** misc *** 1) * Generated man page is not removed on clean. Added debian/clean for that purpose. 2) Probably you don't need this reminder but when convenient please forward patch and man pages to upstream. Yes, all right, I'm going to do that. Everything has been pushed to the Git repo. Greetings, Daniel -- http://www.danielstender.com/blog/ GPG key ID: 1654BD9C -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51c86e2b.90...@danielstender.com
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hi Daniel, One more thing: please install desktop icon. At the moment it is mentioned in .desktop file but not yet shipped by the package. Regards, Dmitry. -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/201306202004.26028.only...@member.fsf.org
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Fixed. Rest is on the way. Greetings, Daniel On 20.06.2013 12:04, Dmitry Smirnov wrote: Hi Daniel, One more thing: please install desktop icon. At the moment it is mentioned in .desktop file but not yet shipped by the package. Regards, Dmitry. -- http://www.danielstender.com/blog/ GPG key ID: 1654BD9C -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51c393df.6010...@danielstender.com
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
On Tue, 18 Jun 2013 23:15:23 Mathieu Malaterre wrote: Technically RelWithDebInfo should not be used anymore with cmake from sid: http://lists.debian.org/debian-devel/2013/06/msg00278.html It now appends -DNDEBUG ... see #701231 for more info Interesting, thanks for this information. I'm just wondering if should not is a little bit too strong to express new recommended practice... If package builds according to expectations with RelWithDebInfo then perhaps it might be safe to keep as is... Best wishes, Dmitry Smirnov. --- Good luck happens when preparedness meets opportunity. -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/201306191946.04651.only...@debian.org
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
On Wed, 19 Jun 2013 03:25:05 Daniel Stender wrote: 1) buildflags.patch / build type First of all, the patch works fine. Actually, after removing the forced variable overrides Cmake recognizes already the standard environment build flags. Great, thanks for checking. :) I've came across that Debhelper 20130504 was set to always switch to RelWithDebInfo build type (#701233), which yes puts -DNDEBUG into building on current Sid (Debhelper 20130605) - are there different opinions on that issue? It is great than debhelper became better with cmake. From my past experience with cmake I might be a bit traumatised by weird upstream scripts that strip executables in Release mode. However explicitly passing build type may be advantages in some cases. It does not hide logic and allows to set proper corresponding CMAKE variables that are specific to build type. Concerning backportability it would be even preferable to explicitly set build type to produce reliable and predictable build results. For scantailor I have no preference and leave decision to maintainer. 2) hardening of scantailor-cli As another issue remains that scantailor-cli still doesn't got fortified: $ hardening-check scantailor-cli As far as I'm aware `hardening-check` might produce incorrect results. When I check build logs with `blhc` it prints no warnings whatsoever which is an evidence that all executables compiled will hardening flags so IMHO we're all good here. Here is another minor improvement suggestion: there are some unnecessary versioned build-dependencies on * cmake (= 2.6.0): oldest is 2.8.2+dfsg.1-0+squeeze1 * libqt4-dev (= 4.4.0): oldest is 4:4.6.3-4+squeeze1 Also I'd set source archive and .DEB files compression to xz: Source compression can be set in debian/source/options by adding compression = xz .DEB file compression can be set in debian/rules by adding override_dh_builddeb: dh_builddeb -- -Zxz I'll have yet another look at the package tonight. Thanks. Cheers, Dmitry Smirnov. -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/201306192009.50634.only...@debian.org
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
All right, On 19.06.2013 12:09, Dmitry Smirnov wrote: Here is another minor improvement suggestion: there are some unnecessary versioned build-dependencies on * cmake (= 2.6.0): oldest is 2.8.2+dfsg.1-0+squeeze1 * libqt4-dev (= 4.4.0): oldest is 4:4.6.3-4+squeeze1 Changed. Also I'd set source archive and .DEB files compression to xz: Source compression can be set in debian/source/options by adding compression = xz .DEB file compression can be set in debian/rules by adding override_dh_builddeb: dh_builddeb -- -Zxz Also added that. I'll have yet another look at the package tonight. Welcome! :-) Greetings, Daniel Stender -- http://www.danielstender.com/blog/ GPG key ID: 1654BD9C -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51c1b37a.7070...@danielstender.com
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hi Daniel, Thanks for your recent corrections. Just few more left to do and I'll upload for you. * When I mentioned unnecessary versioned dependencies I meant that versioning is unnecessary and safe to drop. There is no point writing cmake (= 2.8.2+dfsg.1-0+squeeze1) when just cmake is enough because cmake in oldstable is above minimum requirement. Same applies to libqt4-dev. * Short package description should start with lowercase letter. Copyright: * Upstream uses two emails in source files: Joseph Artsimovich joseph.artsimov...@gmail.com Joseph Artsimovich josep...@mail.ru The latter statement appears to be older (it is often accompanied by years 2007-2008 and one would hardly use this horrible russian email service while having account at gmail). I would at least mention first statement (i.e. contact email) in Upstream-Contact or even include both statement to copyright paragraph as it would be speculative to decide which statement is more important/valid. * The following copyright holders are not listed in debian/copyright: Vadim Kuznetsov ()DikBSD dik...@gmail.com 2011 Petr Kovar pej...@gmail.com * Based on code from the GIMP project is not a copyright statement so perhaps it's better to move it to Comment section. Then it would become obvious that word Copyright is used twice in the paragraph. * At least one file in packaging/osx is under BOOST license. Also file ScanTailor.icns is a (sourceless?) binary file... * License: PD perhaps is better written as License: public-domain. For files in Public-Domain copyright is not applicable so it is better to write Copyright: not-applicable but mention origin of files (i.e. Tango project and URL) in Comment field. * One of the man pages is written by Artem Popov art...@gmail.com but that's not how his name is written in debian/copyright. It is always safer to use original statement as written by author without unnecessary translation of person's name to native characters even if it is done correctly. * Generated man page is not removed on clean. Probably you don't need this reminder but when convenient please forward patch and man pages to upstream. Thanks for your patience. Correcting the above issues will increase our chances to pass ftp-master's review. Cheers, Dmitry Smirnov. -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/201306200642.13719.only...@debian.org
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
On Sat, Jun 15, 2013 at 6:20 PM, Dmitry Smirnov only...@debian.org wrote: Build type is better to leave as RELWITHDEBINFO. This might be useful if you decide to provide -dbg package or just to (re-)build with debugging info with command like Technically RelWithDebInfo should not be used anymore with cmake from sid: http://lists.debian.org/debian-devel/2013/06/msg00278.html It now appends -DNDEBUG ... see #701231 for more info 2cts -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/ca+7wuszxrsb7cpda3aqb9ufvgmauzpoe9ub+2uydp0uujtm...@mail.gmail.com
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
1) buildflags.patch / build type First of all, the patch works fine. Actually, after removing the forced variable overrides Cmake recognizes already the standard environment build flags. I've came across that Debhelper 20130504 was set to always switch to RelWithDebInfo build type (#701233), which yes puts -DNDEBUG into building on current Sid (Debhelper 20130605) - are there different opinions on that issue? Dropping any custom switches (instead of addition of --as-needed linker flag) Scantailor on Sid for now builds with the flags (CPP_FLAGS got injected into C_FLAGS since Debhelper 20120417, #653916): -g -O2 -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2 -O2 -g -DNDEBUG Please further discuss that. 2) hardening of scantailor-cli As another issue remains that scantailor-cli still doesn't got fortified: $ hardening-check scantailor-cli scantailor-cli: Position Independent Executable: no, normal executable! Stack protected: yes Fortify Source functions: no, only unprotected functions found! Read-only relocations: yes Immediate binding: no not found! I'll get deeper into that issue. Latest commit: http://anonscm.debian.org/gitweb/?p=collab-maint/scantailor.git;a=commitdiff;h=e37cf9eecb8223e524958de150a897cfba28a9dd build log: http://www.danielstender.com/uploads/scantailor_0.9.11.1-1_amd64.build Much thanks for comments so far greetings, Daniel Stender On 18.06.2013 15:15, Mathieu Malaterre wrote: On Sat, Jun 15, 2013 at 6:20 PM, Dmitry Smirnov only...@debian.org wrote: Build type is better to leave as RELWITHDEBINFO. This might be useful if you decide to provide -dbg package or just to (re-)build with debugging info with command like Technically RelWithDebInfo should not be used anymore with cmake from sid: http://lists.debian.org/debian-devel/2013/06/msg00278.html It now appends -DNDEBUG ... see #701231 for more info 2cts -- http://www.danielstender.com/blog/ GPG key ID: 1654BD9C -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51c097f1.7010...@danielstender.com
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
As a matter of fact, the CXX_BUILDFLAGS are recognized after a 2nd cmake run: http://www.cmake.org/pipermail/cmake/2013-June/055082.html Greetings, Daniel On 15.06.2013 18:20, Dmitry Smirnov wrote: 3) buildflags But Scantailor gets compiled w/o any customization That's because upstream drop/override given CXXFLAGS (which is a bug). :) -- http://www.danielstender.com/blog/ GPG key ID: 1654BD9C -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51beefbd.20...@danielstender.com
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
On Mon, 17 Jun 2013 21:15:09 Daniel Stender wrote:
As a matter of fact, the CXX_BUILDFLAGS are recognized after a 2nd cmake run:
http://www.cmake.org/pipermail/cmake/2013-June/055082.html
Interesting... I found another post discussing similar bug:
http://www.cmake.org/pipermail/cmake/2008-October/024550.html
Although I don't have good understanding of how it works, using the
above information I was able to produce patch (attached) that fixes
incorrect FLAGS override. I tested it and it looks like no build flags
are dropped any more. I would appreciate if you could review the patch
and apply it if appropriate.
Regards,
Dmitry Smirnov.
---
Platitude: an idea (a) that is admitted to be true by everyone, and (b)
that is not true.
-- H. L. Mencken
Last-Update: 2013-06-18
Forwarded: no
Author: Dmitry Smirnov only...@member.fsf.org
Description: fix FLAGS override due to incorrect caching
Read more about similar issue at
http://www.cmake.org/pipermail/cmake/2008-October/024550.html
--- a/cmake/SetDefaultGccFlags.cmake
+++ b/cmake/SetDefaultGccFlags.cmake
@@ -53,80 +53,80 @@
# Flags common for all build configurations.
SET(
CMAKE_C_FLAGS
-Wall -Wno-unused -ffast-math ${no_inline_dllexport_cflags_}
-CACHE STRING Common C flags for all build configurations. FORCE
+CACHE STRING Common C flags for all build configurations.
)
SET(
CMAKE_CXX_FLAGS ${CMAKE_C_FLAGS} ${stdlibs_shared_static_}
-CACHE STRING Common C++ flags for all build configurations. FORCE
+CACHE STRING Common C++ flags for all build configurations.
)
# Release
SET(
CMAKE_C_FLAGS_RELEASE
-DNDEBUG -O2 ${visibility_cflags_} ${gc_sections_cflags_}
-CACHE STRING C flags for Release builds. FORCE
+CACHE STRING C flags for Release builds.
)
SET(
CMAKE_CXX_FLAGS_RELEASE
-DNDEBUG -O2 ${visibility_cflags_} ${gc_sections_cflags_}
-CACHE STRING C++ flags for Release builds. FORCE
+CACHE STRING C++ flags for Release builds.
)
SET(
CMAKE_EXE_LINKER_FLAGS_RELEASE
${gc_sections_ldflags_} ${dead_strip_ldflags_}
-CACHE STRING Link flags for Release builds. FORCE
+CACHE STRING Link flags for Release builds.
)
# MinSizeRel
SET(
CMAKE_C_FLAGS_MINSIZEREL
-DNDEBUG -Os ${visibility_cflags_} ${gc_sections_cflags_}
-CACHE STRING C flags for MinSizeRel builds. FORCE
+CACHE STRING C flags for MinSizeRel builds.
)
SET(
CMAKE_CXX_FLAGS_MINSIZEREL
-DNDEBUG -Os ${visibility_cflags_} ${gc_sections_cflags_}
-CACHE STRING C++ flags for MinSizeRel builds. FORCE
+CACHE STRING C++ flags for MinSizeRel builds.
)
SET(
CMAKE_EXE_LINKER_FLAGS_MINSIZEREL
${gc_sections_ldflags_} ${dead_strip_ldflags_}
-CACHE STRING Link flags for MinSizeRel builds. FORCE
+CACHE STRING Link flags for MinSizeRel builds.
)
# RelWithDebInfo
SET(
CMAKE_C_FLAGS_RELWITHDEBINFO
-DNDEBUG -g -O2 ${visibility_cflags_}
-CACHE STRING C flags for RelWithDebInfo builds. FORCE
+CACHE STRING C flags for RelWithDebInfo builds.
)
SET(
CMAKE_CXX_FLAGS_RELWITHDEBINFO
--DNDEBUG -g -O2 ${visibility_cflags_}
-CACHE STRING C++ flags for RelWithDebInfo builds. FORCE
+${CMAKE_CXX_FLAGS_RELWITHDEBINFO} -DNDEBUG -g -O2 ${visibility_cflags_}
+CACHE STRING C++ flags for RelWithDebInfo builds.
)
SET(
CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFO
-CACHE STRING Link flags for RelWithDebInfo builds. FORCE
+CACHE STRING Link flags for RelWithDebInfo builds.
)
# Debug
SET(
CMAKE_C_FLAGS_DEBUG -DDEBUG -g CACHE STRING
-C flags for Debug builds. FORCE
+C flags for Debug builds.
)
SET(
CMAKE_CXX_FLAGS_DEBUG -DDEBUG -g CACHE STRING
-C++ flags for Debug builds. FORCE
+C++ flags for Debug builds.
)
SET(
CMAKE_EXE_LINKER_FLAGS_DEBUG
-CACHE STRING Link flags for Debug builds. FORCE
+CACHE STRING Link flags for Debug builds.
)
- SET(COMPILER_FLAGS_OVERRIDDEN YES CACHE INTERNAL FORCE)
+ SET(COMPILER_FLAGS_OVERRIDDEN YES CACHE INTERNAL )
ENDIF(NOT COMPILER_FLAGS_OVERRIDDEN)
ENDIF(CMAKE_COMPILER_IS_GNUCC AND CMAKE_COMPILER_IS_GNUCXX
--- a/cmake/SetDefaultBuildType.cmake
+++ b/cmake/SetDefaultBuildType.cmake
@@ -1,9 +1,9 @@
MACRO(ST_SET_DEFAULT_BUILD_TYPE TYPE_)
IF(NOT CMAKE_BUILD_TYPE AND NOT DEFAULT_BUILD_TYPE_SET)
- SET(DEFAULT_BUILD_TYPE_SET TRUE CACHE INTERNAL FORCE)
+ SET(DEFAULT_BUILD_TYPE_SET TRUE CACHE INTERNAL )
SET(
CMAKE_BUILD_TYPE ${TYPE_} CACHE STRING
- Build type (Release Debug RelWithDebInfo MinSizeRel) FORCE
+ Build type (Release Debug RelWithDebInfo MinSizeRel)
)
ENDIF(NOT CMAKE_BUILD_TYPE AND NOT DEFAULT_BUILD_TYPE_SET)
ENDMACRO(ST_SET_DEFAULT_BUILD_TYPE)
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Hi Dimitry, much thanks for the review of my package. I've proceeded so far: 1) copyright * fixed * regarding copyright, notation Files: ./crash_reporter/google-breakpad/* is a bit strange as there is no need to prefix file paths with ./ * In Copyright: 2006-2008 Google Inc. paragraph license other is incorrect -- it is a BSD-2-clause license. * In Copyright: 2005-2007 Paul Hsieh paragraph license other is incorrect -- it looks like modified BSD-3-clause license, so perhaps it might be better to mark it something like BSD-3-clause(modified) and/or add a comment. Fixed (google-breakpad as BSD-3-clause). 2) lintian-overrides * fixed * Please do not override no-upstream-changelog. Lack of upstream changelog is a genuine (minor) problem isn't it? Silencing legitimate warnings is not a good idea, especially when you do it without corresponding comment about why the warning was suppressed in lintian-overrides. * I: scantailor: unused-override hardening -no-fortify-functions usr/bin/scantailor-cli I've removed the overrides completely. 3) buildflags * buildflags.patch is incorrect because it is hardcoding FLAGS. To make it useful upstream you could modify cmake to avoid overriding CXXFLAGS and then pass build flags like in the following example: override_dh_auto_configure: dh_auto_configure -- \ -DCMAKE_BUILD_TYPE:STRING=RelWithDebInfo \ -DCMAKE_CXX_FLAGS_RELWITHDEBINFO=$(CPPFLAGS) \ -DCMAKE_EXE_LINKER_FLAGS= -Wl,--as-needed $(LDFLAGS) * Passing build flags requires more work as there are still some FLAGS missing, according to `blhc`: CXXFLAGS missing (-fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security) I've removed the patch and created an override for auto configure. Nevertheless, passing the build flags stays problematic. The Cmake invokes with fully expanded buildflag variables (I've added CXXFLAGS): cut cmake .. -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_VERBOSE_MAKEFILE=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo - -DCMAKE_BUILD_TYPE:STRING=RelWithDebInfo -DCMAKE_CXX_FLAGS_RELWITHDEBINFO=-g -O2 - -fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security -D_FORTIFY_SOURCE=2 -DCMAKE_EXE_LINKER_FLAGS= -Wl,--as-needed -Wl,-z,relro /cut But Scantailor gets compiled w/o any customization, running the Cmake standard build flags: cut /usr/lib/ccache/c++ -DBOOST_MULTI_INDEX_DISABLE_SERIALIZATION -DBOOST_TEST_DYN_LINK - -D_BIND_TO_CURRENT_VCLIBS_VERSION=1 -Wall -Wno-unused -ffast-math -DNDEBUG -g -O2 - -fvisibility=hidden -I/tmp/buildd/scantailor-0.9.11.1/. ... /cut Anyway, the custom linker flags got passed through! I've experimented around, e.g. dropping the build type switch, but couldn't got a clue why this isn't working that way, is it? For recent changes, please cf.: http://anonscm.debian.org/gitweb/?p=collab-maint/scantailor.git Build log: http://www.danielstender.com/uploads/scantailor_0.9.11.1-1_amd64.build Mucht thanks also for the other comments! Greetings, Daniel Stender - -- http://www.danielstender.com/blog/ GPG key ID: 1654BD9C -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.11 (GNU/Linux) Comment: Using GnuPG with undefined - http://www.enigmail.net/ iEYEARECAAYFAlG8hUsACgkQU8fCMhZUvZyKqACdGxpJtcEGSTICZV5ZiiieJm2O z4sAn2jSeJUlo0Fy0VYglfqzF9TUEqou =Bo6n -END PGP SIGNATURE- -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51bc854e.2080...@danielstender.com
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hi Daniel,
On Sun, 16 Jun 2013 01:16:30 Daniel Stender wrote:
3) buildflags
But Scantailor gets compiled w/o any customization
That's because upstream drop/override given CXXFLAGS (which is a
bug). :)
I'm not experienced with cmake but perhaps first attempt to fix this
problem might look like the following patch:
--- a/cmake/SetDefaultGccFlags.cmake
+++ b/cmake/SetDefaultGccFlags.cmake
@@ -102,9 +102,9 @@
CACHE STRING C flags for RelWithDebInfo
builds. FORCE
)
SET(
CMAKE_CXX_FLAGS_RELWITHDEBINFO
- -DNDEBUG -g -O2 ${visibility_cflags_}
+ ${CMAKE_CXX_FLAGS_RELWITHDEBINFO} -DNDEBUG
-g -O2 ${visibility_cflags_}
CACHE STRING C++ flags for RelWithDebInfo
builds. FORCE
)
SET(
CMAKE_EXE_LINKER_FLAGS_RELWITHDEBINFO
There are might be better ways to preserve flags and I'm not sure if I
did it properly. Also CFLAGS might need similar workaround (I didn't
check).
Anyway, the custom linker flags got passed through!
Good. :)
I've experimented around, e.g. dropping the build type switch, but couldn't
got a clue why this
isn't working that way, is it?
Build type is better to leave as RELWITHDEBINFO. This might be
useful if you decide to provide -dbg package or just to (re-)build
with debugging info with command like
`DEB_BUILD_OPTIONS=nostrip debuild -us -b`
For recent changes, please cf.:
http://anonscm.debian.org/gitweb/?p=collab-maint/scantailor.git
Sure, I'll have a look in few days (as soon as I have time). Hopefully
it will be ready for upload by then. ;)
Mucht thanks also for the other comments!
No worries, you're very welcome. :)
Cheers,
Dmitry Smirnov
GPG key : 4096R/53968D1B
--
To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/201306160220.01580.only...@debian.org
Re: Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hi, Am Wed, 12 Jun 2013 20:05:38 +0200 schrieb kardan kar...@riseup.net: I tried to compile scantailor and am stuck with zlib1g. Could not find zlib headers. I solved this, by reinstalling the package. Obviously the header file has been lost on the disk. Now there is another issue with boost. $ cmake . -- Boost version: 1.49.0 -- Could NOT find Boost CMake Error at CMakeLists.txt:297 (MESSAGE): Could not find boost headers or libraries. You may need to install a package named libboost1.35-dev or similarly. ii libboost-dev1.49.0.1 The interesting line CMakeLists.txt is 295: FIND_PACKAGE(Boost 1.35.0 COMPONENTS unit_test_framework prg_exec_monitor) /usr/share/cmake-2.8/Modules/FindBoost.cmake # Currently this module searches for the following version numbers: # 1.33, 1.33.0, 1.33.1, 1.34, 1.34.0, 1.34.1, 1.35, 1.35.0, 1.35.1, # 1.36, 1.36.0, 1.36.1, 1.37, 1.37.0, 1.38, 1.38.0, 1.39, 1.39.0, # 1.40, 1.40.0, 1.41, 1.41.0, 1.42, 1.42.0, 1.43, 1.43.0, 1.44, 1.44.0, # 1.45, 1.45.0, 1.46, 1.46.0, 1.46.1, 1.47, 1.47.0, 1.48, 1.48.0, # 1.49, 1.49.0, 1.50, 1.50.0, 1.51, 1.51.0, 1.52, 1.52.0, # 1.53, 1.53.0, 1.54, 1.54.0, 1.55, 1.55.0, 1.56, 1.56.0 With my limited view I assume it is not necessary to request this specific version. Otherwise I see obstacles to use scantailor on wheezy. Any hints are welcome. thanks, kardan -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130614151110.4b2dd014@delight
Re: Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Now there is another issue with boost. $ cmake . -- Boost version: 1.49.0 -- Could NOT find Boost CMake Error at CMakeLists.txt:297 (MESSAGE): Could not find boost headers or libraries. You may need to install a package named libboost1.35-dev or similarly. ii libboost-dev1.49.0.1 The interesting line CMakeLists.txt is 295: FIND_PACKAGE(Boost 1.35.0 COMPONENTS unit_test_framework prg_exec_monitor) /usr/share/cmake-2.8/Modules/FindBoost.cmake # Currently this module searches for the following version numbers: # 1.33, 1.33.0, 1.33.1, 1.34, 1.34.0, 1.34.1, 1.35, 1.35.0, 1.35.1, # 1.36, 1.36.0, 1.36.1, 1.37, 1.37.0, 1.38, 1.38.0, 1.39, 1.39.0, # 1.40, 1.40.0, 1.41, 1.41.0, 1.42, 1.42.0, 1.43, 1.43.0, 1.44, 1.44.0, # 1.45, 1.45.0, 1.46, 1.46.0, 1.46.1, 1.47, 1.47.0, 1.48, 1.48.0, # 1.49, 1.49.0, 1.50, 1.50.0, 1.51, 1.51.0, 1.52, 1.52.0, # 1.53, 1.53.0, 1.54, 1.54.0, 1.55, 1.55.0, 1.56, 1.56.0 With my limited view I assume it is not necessary to request this specific version. Otherwise I see obstacles to use scantailor on wheezy. I found (1) and there under Configuration: * If you have missing dependencies you will get an error message telling you what is missing. You can then search for the missing package and install it. * If you have a library and header files installed in a non-standard place then cmake will not find them. In that case you can run the interactive program ccmake which allows you to specify paths to libraries and header files. My POV is: As you already found the error (line 295), your conclusion is wrong. You have 1.49.0.1, but cmake searches for 1.35. So patch CMakeLists.txt or run ccmake. If you don't want scantailor just to compile and run on your personal Wheezy box, compile for unstable, wait for testing migration and then do a backport. HTH, Martin 1: http://sourceforge.net/apps/mediawiki/scantailor/index.php?title=Building_from_source_code_on_Linux_and_Mac_OS_X -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51bb1f87.6060...@gmx.de
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hi Martin,
thanks for your answer.
Am Fri, 14 Jun 2013 15:49:59 +0200
schrieb Martin Eberhard Schauer martin.e.scha...@gmx.de:
My POV is: As you already found the error (line 295), your conclusion
is wrong. You have 1.49.0.1, but cmake searches for 1.35.
You are right and I am confused. Boost has already been found at the
beginning but this is forgotten later. I learn, that FIND_PACKAGE
overrides all former states even if it was found already.
Also from [1] I learned that FIND_PACKAGE does not fail because of the
higher version as EXACT is not set, but as you can see in the following
it is about the COMPONENTS.
So patch CMakeLists.txt or run ccmake.
$ grep -n -i boost CMakeLists.txt
74:INCLUDE(FindBoost)
152:FILE(GLOB boost_dirs1 ${build_outer_dir}/boost_1_[0-9]*_[0-9]*)
153:FILE(GLOB boost_dirs2 ${source_outer_dir}/boost_1_[0-9]*_[0-9]*)
286:BOOST_ROOT boost-build.jam PATHS ${boost_dirs1} ${boost_dirs2}
287:DOC Path to top-level Boost source directory.
289:SET(Boost_USE_STATIC_LIBS ON)
291:ADD_DEFINITIONS(-DBOOST_TEST_DYN_LINK)
294:SET(Boost_USE_MULTITHREADED ON)
295:FIND_PACKAGE(Boost 1.35.0 COMPONENTS unit_test_framework prg_exec_monitor)
296:FIND_PACKAGE(Boost 1.49.0 COMPONENTS unit_test_framework prg_exec_monitor)
297:FIND_PACKAGE(Boost 1.49.0)
298:FIND_PACKAGE(Boost 1.49.1 COMPONENTS unit_test_framework prg_exec_monitor)
299:IF(NOT Boost_FOUND)
302:Could not find boost headers or libraries.\n
303:You may need to install a package named libboost1.35-dev or
later.
305:ENDIF(NOT Boost_FOUND)
307:INCLUDE_DIRECTORIES(${Boost_INCLUDE_DIRS})
308:LINK_DIRECTORIES(${Boost_LIBRARY_DIRS})
354:${QT_DEFINITIONS} -DBOOST_MULTI_INDEX_DISABLE_SERIALIZATION
573:# There seems to be a bug in either gcc or boost that makes page
sorting either crash
$ cmake .
-- Boost version: 1.49.0
-- Could NOT find Boost
-- Could NOT find Boost
-- Boost version: 1.49.0
-- Could NOT find Boost
CMake Error at CMakeLists.txt:300 (MESSAGE):
Could not find boost headers or libraries.
You may need to install a package named libboost1.35-dev or later
Only the third line hits (except FindBoost in 74). This leads to the
conclusion the there is some problem with the components
'unit_test_framework prg_exec_monitor'. Please correct me.
If you don't want scantailor just to compile and run on your personal
Wheezy box, compile for unstable, wait for testing migration and then
do a backport.
HTH, Martin
I would like to help. In this case the package is already RFS and I
think repackaging it would be counterproductive?
thanks,
kardan
--
To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20130614164259.2cdc67cc@delight
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hi Kardan, I should have looked at 712056 before my answer. But as I did some C programming in the past, I thought that I understood the problem and Hey, this one is easy to solve :-( I would like to help. In this case the package is already RFS and I think repackaging it would be counterproductive? Yup. There is no need to do it again. But I suggest another point of view. Daniel managed to overcome your problem. His package was reviewed (1) and some work still has to be done. Comparing what you did and what Daniel did may be quite instructive. But instead of browsing in different sources and comparing them it's probably more effictive to have a look at the maint-guide(-xx) (2) and developers-reference (3) packages - in that order. Take it as an exercise. Reading a text book or listening to some lecture is one thing. Making use of some theory in real life *and failing to do so* is the other thing. And that is really instructive :-( Martin 1: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=712056#15 2: http://packages.debian.org/search?suite=defaultsection=allarch=anysearchon=nameskeywords=maint-guide 3: http://packages.debian.org/search?suite=defaultsection=allarch=anysearchon=allkeywords=developer+reference -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51bb3ee7.8010...@gmx.de
Re: Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
My POV is: As you already found the error (line 295), your conclusion
is wrong. You have 1.49.0.1, but cmake searches for 1.35.
My POV is wrong :( (see below)
$ grep -n -i boost CMakeLists.txt
74:INCLUDE(FindBoost)
152:FILE(GLOB boost_dirs1 ${build_outer_dir}/boost_1_[0-9]*_[0-9]*)
153:FILE(GLOB boost_dirs2 ${source_outer_dir}/boost_1_[0-9]*_[0-9]*)
286:BOOST_ROOT boost-build.jam PATHS ${boost_dirs1}
${boost_dirs2}
287:DOC Path to top-level Boost source directory.
289:SET(Boost_USE_STATIC_LIBS ON)
291:ADD_DEFINITIONS(-DBOOST_TEST_DYN_LINK)
294:SET(Boost_USE_MULTITHREADED ON)
295:FIND_PACKAGE(Boost 1.35.0 COMPONENTS unit_test_framework
prg_exec_monitor)
296:FIND_PACKAGE(Boost 1.49.0 COMPONENTS unit_test_framework
prg_exec_monitor)
297:FIND_PACKAGE(Boost 1.49.0)
298:FIND_PACKAGE(Boost 1.49.1 COMPONENTS unit_test_framework
prg_exec_monitor)
299:IF(NOT Boost_FOUND)
302:Could not find boost headers or libraries.\n
303:You may need to install a package named
libboost1.35-dev or later.
305:ENDIF(NOT Boost_FOUND)
You are close here - but I still disagree slightly. As I understand things,
the problem is the incompatibility between the upstream code (above) and the
Debian version - 1.49.0 vs 1.49.0.1.
You are right and I am confused. Boost has already been found at the
beginning but this is forgotten later. I learn, that FIND_PACKAGE
overrides all former states even if it was found already.
Also from [1] I learned that FIND_PACKAGE does not fail because of the
higher version as EXACT is not set, but as you can see in the following
it is about the COMPONENTS.
$ cmake .
-- Boost version: 1.49.0
-- Could NOT find Boost
-- Could NOT find Boost
-- Boost version: 1.49.0
-- Could NOT find Boost
CMake Error at CMakeLists.txt:300 (MESSAGE):
Could not find boost headers or libraries.
but this is forgotten later. Take it as nitpicking - there is a difference
between not used and forgotten. But the *effect* is the same - FTBFS.
Martin
--
To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/51bb467a.6010...@gmx.de
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hi, Am Fri, 14 Jun 2013 18:03:51 +0200 schrieb Martin Eberhard Schauer martin.e.scha...@gmx.de: I should have looked at 712056 before my answer. But as I did some C programming in the past, I thought that I understood the problem and Hey, this one is easy to solve :-( 1: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=712056#15 It's all fine for me now. The problems in the linked comment are still relevant but not in relation to boost. I found out that COMPONENT unit_test_framework is delivered by libboost-test-dev, which was missing. I added a patch to notify the user. 300c300 You may need to install a package named libboost1.35-dev or similarly. --- You may need to install a package named libboost-dev and libboost-test-dev version 1.35 or later. Daniel managed to overcome your problem. His package was reviewed (1) and some work still has to be done. Comparing what you did and what Daniel did may be quite instructive. But instead of browsing in different sources and comparing them it's probably more effictive to have a look at the maint-guide(-xx) (2) and developers-reference (3) packages - in that order. 2: http://packages.debian.org/search?suite=defaultsection=allarch=anysearchon=nameskeywords=maint-guide 3: http://packages.debian.org/search?suite=defaultsection=allarch=anysearchon=allkeywords=developer+reference Thanks. Will use this in the future. Great software! Definitively worth packaging. A first taste (some scans are in poor quality due to the scan): http://heahdk.net/~kardan/books/ The html is created with scantailor2html, find it attached, free for further distribution. Cheers, Kardan300c300 You may need to install a package named libboost1.35-dev or similarly. --- You may need to install a package named libboost-dev and libboost-test-dev version 1.35 or later. scantailor2html Description: Binary data
Re: Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hey Martin,
one last thing from me on this. Thanks for the fun and the help!
the problem is the incompatibility between the upstream code (above)
and the Debian version - 1.49.0 vs 1.49.0.1.
Thanks to the maintainer everything is fine with boost (for this).
Am Fri, 14 Jun 2013 18:36:10 +0200
schrieb Martin Eberhard Schauer martin.e.scha...@gmx.de:
My POV is: As you already found the error (line 295), your
conclusion is wrong. You have 1.49.0.1, but cmake searches for
1.35.
My POV is wrong :( (see below)
$ grep -n -i boost CMakeLists.txt
74:INCLUDE(FindBoost)
152:FILE(GLOB boost_dirs1
${build_outer_dir}/boost_1_[0-9]*_[0-9]*) 153:FILE(GLOB
boost_dirs2 ${source_outer_dir}/boost_1_[0-9]*_[0-9]*)
286:BOOST_ROOT boost-build.jam PATHS ${boost_dirs1}
${boost_dirs2}
287:DOC Path to top-level Boost source directory.
289:SET(Boost_USE_STATIC_LIBS ON)
291:ADD_DEFINITIONS(-DBOOST_TEST_DYN_LINK)
294:SET(Boost_USE_MULTITHREADED ON)
295:FIND_PACKAGE(Boost 1.35.0 COMPONENTS unit_test_framework
prg_exec_monitor)
296:FIND_PACKAGE(Boost 1.49.0 COMPONENTS unit_test_framework
prg_exec_monitor)
297:FIND_PACKAGE(Boost 1.49.0)
298:FIND_PACKAGE(Boost 1.49.1 COMPONENTS unit_test_framework
prg_exec_monitor)
299:IF(NOT Boost_FOUND)
302:Could not find boost headers or libraries.\n
303:You may need to install a package named
libboost1.35-dev or later.
305:ENDIF(NOT Boost_FOUND)
You are close here - but I still disagree slightly. As I understand
things, the problem is the incompatibility between the upstream code
(above) and the Debian version - 1.49.0 vs 1.49.0.1.
Boost has already been found at
the beginning but this is forgotten later.
$ cmake .
-- Boost version: 1.49.0
-- Could NOT find Boost
-- Could NOT find Boost
-- Boost version: 1.49.0
-- Could NOT find Boost
CMake Error at CMakeLists.txt:300 (MESSAGE):
Could not find boost headers or libraries.
but this is forgotten later. Take it as nitpicking - there is a
difference between not used and forgotten. But the *effect* is
the same - FTBFS.
(: nitpicking back: The error is raised because of the last
unsuccessfull FIND_PACKAGE(Boost 1.49.1 COMPONENTS unit_test_framework
prg_exec_monitor) which fails because neither 1.49.1 nor
libboost-test-dev are installed and overrides the success of the former
test (from libboost-dev package info: Version: 1.49.0.1, This package
is a dependency package, which depends on Debian's default Boost
version (currently 1.49)).
FIND_PACKAGE(Boost 1.49.0 COMPONENTS unit_test_framework
prg_exec_monitor)
FIND_PACKAGE(Boost 1.49.1)
-- Boost version: 1.49.0
-- Found the following Boost libraries:
-- unit_test_framework
-- prg_exec_monitor
-- Could NOT find Boost
CMake Error at CMakeLists.txt:298 (MESSAGE):
Could not find boost headers or libraries.
You may need to install packages named libboost-dev and
libboost-test-dev version 1.35 or later.
I think the topic is ready to bury :)
Kardan
--
To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20130614195618.760500e0@delight
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Package: sponsorship-requests Severity: wishlist Dear mentors, I am looking for a sponsor for my initial package of Scantailor 0.9.11.1 for Debian. I've uploaded it to mentors: http://mentors.debian.net/package/scantailor dget -x http://mentors.debian.net/debian/pool/main/s/scantailor/scantailor_0.9.11.1-1.dsc and created a Git repo belonging to the collab-maint group: http://anonscm.debian.org/gitweb/?p=collab-maint/scantailor.git I've builded against Sid using Cowbuilder, I've uploaded my build log here: http://www.danielstender.com/uploads/scantailor_0.9.11.1-1_amd64.build Info on the program could be found here: http://scantailor.sourceforge.net/ Thanks for any hint, pointers, and reviews in advance! Regards, Daniel Stender -- http://www.danielstender.com/blog/ GPG key ID: 1654BD9C -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/51b8975e.1030...@danielstender.com
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Hi Daniel, I tried to compile scantailor and am stuck with zlib1g. $ cmake . -- Boost version: 1.49.0 CMake Error at CMakeLists.txt:193 (MESSAGE): Could not find zlib headers. You may need to install a package named zlib1g-dev or similarly. ii zlib1g:i386 1:1.2.7.dfsg-13 ii zlib1g-dbg 1:1.2.7.dfsg-13 ii zlib1g-dev:i386 1:1.2.7.dfsg-13 Description:Debian GNU/Linux 7.0 (wheezy) Release:7.0 Do you have an idea? thanks, kardan Am Wed, 12 Jun 2013 17:44:30 +0200 schrieb Daniel Stender dan...@danielstender.com: Package: sponsorship-requests Severity: wishlist Dear mentors, I am looking for a sponsor for my initial package of Scantailor 0.9.11.1 for Debian. I've uploaded it to mentors: http://mentors.debian.net/package/scantailor dget -x http://mentors.debian.net/debian/pool/main/s/scantailor/scantailor_0.9.11.1-1.dsc and created a Git repo belonging to the collab-maint group: http://anonscm.debian.org/gitweb/?p=collab-maint/scantailor.git I've builded against Sid using Cowbuilder, I've uploaded my build log here: http://www.danielstender.com/uploads/scantailor_0.9.11.1-1_amd64.build Info on the program could be found here: http://scantailor.sourceforge.net/ Thanks for any hint, pointers, and reviews in advance! Regards, Daniel Stender -- To UNSUBSCRIBE, email to debian-mentors-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20130612200538.52b0a8dc@delight
Bug#712056: RFS: scantailor [ITP] -- interactive post-processing tool for scanned document pages
Control: tags -1 moreinfo Hi Daniel, Thanks for packaging scantailor. I had a brief look at the package and IMHO it needs a little bit more work before we'll be able to upload it. * buildflags.patch is incorrect because it is hardcoding FLAGS. To make it useful upstream you could modify cmake to avoid overriding CXXFLAGS and then pass build flags like in the following example: override_dh_auto_configure: dh_auto_configure -- \ -DCMAKE_BUILD_TYPE:STRING=RelWithDebInfo \ -DCMAKE_CXX_FLAGS_RELWITHDEBINFO=$(CPPFLAGS) \ -DCMAKE_EXE_LINKER_FLAGS= -Wl,--as-needed $(LDFLAGS) Also perhaps more DEP-3 headers like [Forwarded,Last-Update] may be useful. * Passing build flags requires more work as there are still some FLAGS missing, according to `blhc`: CXXFLAGS missing (-fstack-protector --param=ssp-buffer-size=4 -Wformat -Werror=format-security) * I: scantailor: unused-override hardening-no-fortify-functions usr/bin/scantailor-cli * Please do not override no-upstream-changelog. Lack of upstream changelog is a genuine (minor) problem isn't it? Silencing legitimate warnings is not a good idea, especially when you do it without corresponding comment about why the warning was suppressed in lintian-overrides. * regarding copyright, notation Files: ./crash_reporter/google-breakpad/* is a bit strange as there is no need to prefix file paths with ./ * In Copyright: 2006-2008 Google Inc. paragraph license other is incorrect -- it is a BSD-2-clause license. * In Copyright: 2005-2007 Paul Hsieh paragraph license other is incorrect -- it looks like modified BSD-3-clause license, so perhaps it might be better to mark it something like BSD-3-clause(modified) and/or add a comment. All the best, Dmitry Smirnov GPG key : 4096R/53968D1B --- No person, no idea, and no religion deserves to be illegal to insult, not even the Church of Emacs. -- Richard Stallman signature.asc Description: This is a digitally signed message part.
