For upstream #121
For upstream #121
Conversation
The current version includes updated autotools/libtool files, which now are confused by the extra $(SHELL) stuff that was in the build rules.
ghost
assigned Runtime tests have revealed that libMagick++ crashes in pthread-related code. Therefore, this patch disables threading in the package. Unfortunately the configure switch is not enough; the resulting config header needs to be tweaked.
|
Here is one additional patch for imagemagick. This fixes runtime crashes in my application. Please consider. |
|
The shell specifications in ilmbase were there to overcome some bashisms - hopefully they have been removed, but I don't have a machine to test with at the moment. Does the native build of ilmbase in openexr also need to be changed? Alternately, are the changes really necessary in the cross-build of ilmbase - why would one work and not the other? |
|
Good point. I rechecked, and there is another way to fix the problem. First, with the original ilmbase.mk, I get a libIlmThread.a that contains only a few bytes, and linking against this fails (see the openexr linking test in the merge request). The difference between the ilmbase builds in ilmbase.mk and openexr.mk is that ilmbase.mk has CONFIG_SHELL=$(SHELL) at the end of configure, whereas openexr.mk has additionally SHELL=$(SHELL). Putting this SHELL=$(SHELL) also in ilmbase.mk is an alternative way to fix the problem, and probably also avoids the bashisms. So instead of 3ce209e, which removes all The other patches on top of that still work. |
|
Great! I've cherry-picked the three other commits and added the extra SHELL variable 44441d4 |
These patches fix package ilmbase, on which openexr is based, and add a link test to openexr. Based on these changes, a fixed imagemagick package now also has a link test.