| 1 |
TL/DR: Skip to "WORK-AROUND". |
| 2 |
|
| 3 |
PROBLEM |
| 4 |
======= |
| 5 |
|
| 6 |
Unfortunately, FTGL has contained two strictly incompatible |
| 7 |
behaviours for a long time: |
| 8 |
|
| 9 |
The initial library set some OpenGL state, e.g. glEnable(GL_BLEND) |
| 10 |
and glBlendFunc(), from the rendering functions. Not only may this |
| 11 |
not be what the user wants, the blending function used is also not |
| 12 |
quite correct. See |
| 13 |
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=742469) for a |
| 14 |
detailed explanation and example. |
| 15 |
|
| 16 |
A fix made by sammy in 2009, see |
| 17 |
https://github.com/frankheckenbach/ftgl/commit/29603ae3fa88c5b9e079a6db23be2cdea95aef39 |
| 18 |
removed these settings, but it was never picked up by Debian and |
| 19 |
possibly other distributions. |
| 20 |
|
| 21 |
As a result, programs assuming the new behaviour would behave |
| 22 |
wrongly without the fix. This cannot be changed from within those |
| 23 |
programs, i.e. outside of the library, because the library without |
| 24 |
the fix would override the settings. |
| 25 |
|
| 26 |
OTOH, programs built against the original version would assume these |
| 27 |
settings and not render correctly with the fix applied. These could |
| 28 |
be fixed in principle by making the settings from within the |
| 29 |
programs, but after almost 10 years of ignoring the problem, some of |
| 30 |
those programs have apparently become unmaintained, or maintained |
| 31 |
just to the bare minimum, so expecting them all to fix the issue |
| 32 |
seems unrealistic. |
| 33 |
|
| 34 |
So as a kludge, version 2.4 of FTGL adds a new flag to choose the |
| 35 |
behaviour, as described under "WORK-AROUND" (not a real solution, |
| 36 |
but the best we can realistically get). |
| 37 |
|
| 38 |
To avoid mismatch between the two behaviours, the library will only |
| 39 |
allow to set this flag to the same value (but any number of times) |
| 40 |
within one program run. So, e.g. if your code includes some |
| 41 |
libraries that in turn use FTGL, and they all set this flag to |
| 42 |
false, all is fine. But if some of them set it to false, and some to |
| 43 |
true, things can't work, as described above. To avoid confusion, on |
| 44 |
an attempt to set the flag to a different value than was set before, |
| 45 |
FTGL will throw an exception (a std::logic_error which should not be |
| 46 |
caught in general usually, neither here in particular). |
| 47 |
|
| 48 |
|
| 49 |
WORK-AROUND |
| 50 |
=========== |
| 51 |
|
| 52 |
The handling of OpenGL state should now be set with |
| 53 |
|
| 54 |
FTLibrary::Instance().LegacyOpenGLState(On); |
| 55 |
|
| 56 |
You should add a call to this function near the start of all your |
| 57 |
programs that use FTGL, and in some strategic place in all libraries |
| 58 |
that use FTGL. |
| 59 |
|
| 60 |
If your code relies on FTGL automatically setting some OpenGL state, |
| 61 |
like the versions in Debian and possibly other distributions do, |
| 62 |
pass "true" for the "On" parameter. If you don't call |
| 63 |
LegacyOpenGLState, this is the default behaviour for now, but in the |
| 64 |
future, FTGL will warn about it and finally fail. |
| 65 |
|
| 66 |
If you need manual control over the OpenGL state, like the versions |
| 67 |
on GitHub do, pass "false". |
| 68 |
|
| 69 |
Even if you use the legacy version ("true"), you might want to |
| 70 |
consider changing to manual setting ("false"). To retain the old |
| 71 |
behaviour then, make the following OpenGL calls from your code, |
| 72 |
either globally, or before each respective rendering, possibly |
| 73 |
protected with glPushAttrib/glPopAttrib: |
| 74 |
|
| 75 |
glEnable(GL_BLEND); |
| 76 |
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA); |
| 77 |
|
| 78 |
To get a better rendering function, instead add: |
| 79 |
|
| 80 |
#define GL_GLEXT_PROTOTYPES |
| 81 |
|
| 82 |
before including OpenGL headers, and then do: |
| 83 |
|
| 84 |
glEnable(GL_BLEND); |
| 85 |
glBlendFuncSeparate(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA, |
| 86 |
GL_ONE, GL_ONE_MINUS_SRC_ALPHA); |