On 5/12/16, 7:34 AM, Sergey Bylokhov wrote:
On 12.05.16 17:23, Philip Race wrote:
Probably we should skip the loading of glib >= 2.2 <=2.29 and bump the
minumum version to 2.20/2.24? This will eliminate the possibility of
the crash.
Not sure what you mean by "skip loading". Where does 2.29 come into
this ?
typo: I meant 2.20
And glib_check_version is available only from 2.6 onward.
A crash will still happen on 2.20 with multiple calls.
Do you mean skip all checking and just "require" 2.24
Yes, because otherwise it will crash if someone will try to load this
library. And I assume that we remove this API because all our
supported platforms support >=2.24, isn't?
It is removed because JavaFX no longer uses it and FX will require a higher
base supported version of OEL/RHEL than Java SE.
But it is still useful in determining if some other random piece of code
loads GTK
with GLIB 2.20 or 2.22 for example.
So we are (slightly) more likely to crash if I remove this code than if
I leave it.
I certainly don't see how removing it *prevents* all crashes.
Think of this fix as really just being about the removal of GThreadHelper
since FX was the only one who used it and it no longer uses it and not
about implicitly bumping the Java SE requirements.
That can be done later if things become more clear.
Is that OK ?
-phil.
I don't see how that would eliminate crashes, and it just increases
the likelihood of us not running at all on more platforms.
-phil.
