This change adds support for ancient macOS lacking libproc. It also
eliminates a bug where exiv2 could not be built on case-sensitive
file systems (the header file is properly TargetConditionals.h).
* avoid re-declaration of constant variables
* Replace pthreads critical section with std::mutex
* ci - better naming
* cmake - increase minimum version to 3.11. Use project DESCRIPTION
* fix - do not treat string::find() return type as bool
* remove conditions that were always true
* remove condition that were always false
* Remove EXV_HAVE_GMTIME_R which is not used anymore
* pixelWidth_ was inherited from Exiv2::Image
The width & height variables in the TiffImage class need to be mutable
to be able to change their values on the getters pixelHeight() and
pixelWidth() ... Do not ask me why ...
* Remove superfluous if
* pvs:V766 item with identical key added already
* pvs:V730 not all members were initialized (time)
* pvs:V730 not all members are initialized
* pvs:v668 no point in testing pointer against null after new
* pvs:V1048 variable assigned the same value
* replace c-style dynamic vector with std one
* pvs:547 fakeData is always true
* Remove useless constructor in derived class
* pvs:V690 modern way to disable copy-ctor
* Replace malloc/free with new/delete. No need to check for null
* pvs:V1028 cast operands and not result
* Remove custom MIN/MAX functions
* pvs:V595 pointer used before verified against null
* pvs: index used before being checked
* pvs:V1028 possible overflow. Cast operands
* pvs:v575 potential null pointer passed to other functions
* pvs:V547 deal with always true/false expressions
* pvs:V560 part of conditional expressions always false or true
* pvs:V701 possible break in realloc -> move to std::vector
* Make some classes 'final'
* Replace sprintf with std::to_string()
* fix compilation on windows
Found with modernize-loop-convert
Ran through git clang-format.
Also removed several questionable loops and replaced with simpler
algorithms.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
I'm a little uneasy about changes to the API of the shared object/DLL. I will investigate #890 during the release process. So I have a plan to identify this if it's an issue. I think it's OK.
```
703 rmills@rmillsmm-local:~/gnu/github/exiv2/bmff/build $ nm --demangle lib/libexiv2.dylib | grep base64
00000000000384a0 T Exiv2::base64decode(char const*, char*, unsigned long)
0000000000038340 T Exiv2::base64encode(void const*, unsigned long, char*, unsigned long)
00000000001c33a0 s Exiv2::base64_encode
704 rmills@rmillsmm-local:~/gnu/github/exiv2/bmff/build $ nm -g --demangle lib/libexiv2.dylib | grep base64
00000000000384a0 T Exiv2::base64decode(char const*, char*, unsigned long)
0000000000038340 T Exiv2::base64encode(void const*, unsigned long, char*, unsigned long)
705 rmills@rmillsmm-local:~/gnu/github/exiv2/bmff/build $
```
Only GNU libc has implementation of the GNU variant of strerror_r, so
also for __GLIBC__ to determine which strerror_r implemenation to use.
This fixes build with musl libc, which makes some, but not all, GNU
extensions available with _GNU_SOURCE.
While taking an EnVar as the parameter is more clear it has the
disadvantage, that passing anything outside of the range of the
enumeration is undefined behavior. The compiler could then optimize
the range check in getEnv away (perfectly legal due to UB), leading
to buffer overreads.
Luis Díaz Más
Evan Miller
Matthias Kuhn
Rosen Penev
Robin Mills
Ingo Weyrich
Luis Díaz Más
Natanael Copa
Dan Čermák