Re: [Bug-ocrad] Version 0.21-pre1 of GNU Ocrad released

[Antonio Diaz Diaz]

Hello Tilman,

Tilman Hausherr wrote:
> - The crashes seem to be solved, i.e. the problem-images don't crash. I
> did not look into the source to check whether you used my "symptom"
> fixes or made your own or both, because you certainly know what you're
> doing.

I have really fixed them, not only the symptoms. Of course, some of 
those fixes can be further improved.


> Another change you might want to make is to start to eliminate some
> warnings:

What warnings?
---------------------------
$ ./configure CXXFLAGS='-Wall -W -O2 --pedantic -Wno-long-long'

creating config.status
creating Makefile
VPATH = .
prefix = /usr/local
exec_prefix = $(prefix)
bindir = $(exec_prefix)/bin
datadir = $(prefix)/share
includedir = ${prefix}/include
infodir = $(datadir)/info
libdir = ${exec_prefix}/lib
mandir = $(datadir)/man
sysconfdir = $(prefix)/etc
CXX = g++
CPPFLAGS =
CXXFLAGS = -Wall -W -O2 --pedantic -Wno-long-long
LDFLAGS =
OK. Now you can run make.
$ make check 2>&1 | tee x
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o common.o common.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o mask.o mask.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o rational.o rational.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o rectangle.o rectangle.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o track.o track.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o ucs.o ucs.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o page_image.o page_image.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o page_image_io.o 
page_image_io.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o bitmap.o bitmap.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o blob.o blob.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o profile.o profile.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o feats.o feats.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o feats_test0.o 
feats_test0.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o feats_test1.o 
feats_test1.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o character.o character.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o character_r11.o 
character_r11.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o character_r12.o 
character_r12.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o character_r13.o 
character_r13.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o textline.o textline.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o textline_r2.o 
textline_r2.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o textblock.o textblock.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o textpage.o textpage.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o arg_parser.o arg_parser.cc
g++  -Wall -W -O2 --pedantic -Wno-long-long -DPROGVERSION=\"0.21-pre1\" 
-c -o main.o main.cc
g++  -o ocrad common.o mask.o rational.o rectangle.o track.o ucs.o 
page_image.o page_image_io.o bitmap.o blob.o profile.o feats.o 
feats_test0.o feats_test1.o character.o character_r11.o character_r12.o 
character_r13.o textline.o textline_r2.o textblock.o textpage.o 
arg_parser.o main.o
g++  -Wall -W -O2 --pedantic -Wno-long-long -c -o ocradlib.o ocradlib.cc
ar -rcs libocrad.a common.o mask.o rational.o rectangle.o track.o ucs.o 
page_image.o page_image_io.o bitmap.o blob.o profile.o feats.o 
feats_test0.o feats_test1.o character.o character_r11.o character_r12.o 
character_r13.o textline.o textline_r2.o textblock.o textpage.o ocradlib.o
g++  -Wall -W -O2 --pedantic -Wno-long-long -DPROGVERSION=\"0.21-pre1\" 
-c -o ocrcheck.o ocrcheck.cc
g++  -o ocrcheck ocrcheck.o libocrad.a
testing 
ocrad-0.21-pre1.........................................................................................................................................
tests completed successfully.
$
---------------------------


> replace
>         const double cmom_max = cmom[maxval];
> with
>         const double cmom_max = (double) cmom[maxval];

Didn't the authors of your compiler ever heard about implicit type 
conversion?
http://en.wikipedia.org/wiki/Type_conversion#Implicit_type_conversion


> there are more warnings (especially when compiling for 64bit, because
> you're not using size_t type for numbers based on sizes), but this is a
> small start.

No, this is the end. I do not plan to use size_t, off_t and the like 
anytime soon. In page_image_io.cc (line 247) I verify that the number of 
pixels in the image is no larger than INT_MAX and then I use int for all 
sizes, avoiding more conversion issues while being sure that overflows 
won't happen.

On projects where 32 bit sizes are not enough I simply use long long 
instead of int.


> While most warnings are just that, they "mask" warnings
> that might really have a meaning, and they're scary for newbies.

Then, please, show me the warnings with real meaning. About the newbies, 
I hope they get scared enough as to migrate to GCC or even to GNU/Linux. :-)


Regards,
Antonio.