It's also fix bug with missing plots, when used eps files, in octave backend (files eps are exist, but don't show in cantor). This bug observed at least from 15.12 to nowdays (i just always have been forgoting to report about this bug).

hmmm, this is weird, i don't really need this, the current code works for me

MESSAGE("LibSpectre_FOUND: ${LibSpectre_FOUND}")

just after the find_package says
LibSpectre_FOUND: TRUE
here when running on a clean build dir.

But it is true that ${LIBSPECTRE_FOUND} also prints true.

About the add_definitions you should not need it, the generated
src/lib/config-cantorlib.h
already has that define, just make sure you include that file wherever you need it.

Maybe my cmake is newer? I'm on cmake 3.10.3

In D11967#240772, @aacid wrote:

About the add_definitions you should not need it, the generated
src/lib/config-cantorlib.h
already has that define, just make sure you include that file wherever you need it.

Interesting, because add_definitions(-DWITH_EPS) make difference for me (cmake 3.5.1): without this line plots don't appear in cantor (and with this option they do). However LibSpectre_FOUND and LIBSPECTRE_FOUND don't make any difference for me too.

In D11967#240796, @sirgienko wrote:

In D11967#240772, @aacid wrote:

About the add_definitions you should not need it, the generated
src/lib/config-cantorlib.h
already has that define, just make sure you include that file wherever you need it.

Interesting, because add_definitions(-DWITH_EPS) make difference for me (cmake 3.5.1): without this line plots don't appear in cantor (and with this option they do). However LibSpectre_FOUND and LIBSPECTRE_FOUND don't make any difference for me too.

This is probably because there is a fine that uses #if WITH_EPS or #ifdef WITH_EPS and does not include config-cantorlib.h

Probably this? https://paste.kde.org/pnz15dn0w

In D11967#240772, @aacid wrote:

hmmm, this is weird, i don't really need this, the current code works for me

MESSAGE("LibSpectre_FOUND: ${LibSpectre_FOUND}")

just after the find_package says
LibSpectre_FOUND: TRUE
here when running on a clean build dir.

But it is true that ${LIBSPECTRE_FOUND} also prints true.

This is really strange. I don't see nowhere in cantor/cmake/FindLibSpectre.cmake (not sure this file is really used by cmake) the variable LibSpectre_FOUND. Since cmake's variables are case sensitive I'd assume LIBSPECTRE_FOUND is the only correct one. Your cmake is newer than the version that I'm using. But this worked in cantor already in the past for sure, also with older versions of cmake...

About the add_definitions you should not need it, the generated
src/lib/config-cantorlib.h
already has that define, just make sure you include that file wherever you need it.

I don't see any includes for config-cantrolib.h in the code. Maybe this is added somewhere on the cmake-level. Some parts of the current code are aware of the definition of WITH_EPS, e.g. Expression::setResult(). Some other parts like Worksheet:: loginToSession() don't "see" this variable and require now either that include you mentioned or a preprocessor definition. Since #ifdef WITH_EPS is used in several different places in Cantor's code it will take some time now to test all those places. Given the current timeline/deadline for 18.04 I'd rather go with a preprocessor definition via add_defitions() - this is safer now and also more clean and clear, at least to me, since it's straight -forward to see where it's defined.

About the add_definitions you should not need it, the generated
src/lib/config-cantorlib.h
already has that define, just make sure you include that file wherever you need it.

I don't see any includes for config-cantrolib.h in the code. Maybe this is added somewhere on the cmake-level. Some parts of the current code are aware of the definition of WITH_EPS, e.g. Expression::setResult(). Some other parts like Worksheet:: loginToSession() don't "see" this variable and require now either that include you mentioned or a preprocessor definition. Since #ifdef WITH_EPS is used in several different places in Cantor's code it will take some time now to test all those places. Given the current timeline/deadline for 18.04 I'd rather go with a preprocessor definition via add_defitions() - this is safer now and also more clean and clear, at least to me, since it's straight -forward to see where it's defined.

Correction: I was looking for wrong file name. config-cantorlib.h is included at several places. Looks like we simply miss #include <config-cantorlib.h> in wokrsheet.cpp. @sirgienko can you add this include and to try out your test scenarios for octave? It should work also without that add_definiton() line in CMakeLists.

In D11967#241032, @asemke wrote:

Correction: I was looking for wrong file name. config-cantorlib.h is included at several places. Looks like we simply miss #include <config-cantorlib.h> in wokrsheet.cpp. @sirgienko can you add this include and to try out your test scenarios for octave? It should work also without that add_definiton() line in CMakeLists.

Well, It's don't work for me. Plots don't show. But maxima backend starts produce LaTeX output, that's great.

In D11967#240853, @aacid wrote:

Probably this? https://paste.kde.org/pnz15dn0w

Yes, it works, the plots are displayed.

@asemke, Could you please add aacid changes to your diff?

In D11967#241109, @sirgienko wrote:

@asemke, Could you please add aacid changes to your diff?

Done.