search

ReadyTalk / avian

[INACTIVE] Avian is a lightweight virtual machine and class library designed to provide a useful subset of Java's features, suitable for building self-contained applications.
https://readytalk.github.io/avian/
Other
1.22k stars 173 forks source link

fix openjdk/openjdk-src builds for latest JDKs 7 and 8 #472

Closed dicej closed 8 years ago

dicej commented 8 years ago

The tower of patches and hacks grows higher. Ideally, we'll just drop support for JDK 7 soon and clean this mess up a bit, but TravisCI still hasn't gotten the memo that it's dead, so we muddle onward.

I've tested this on Windows, but not yet Linux or OS X. Wanted to get a PR before I move on to that.

dicej commented 8 years ago

I've tested this on Mac, Linux, and Windows. Everything looks fine except on Mac, where the openjdk-src build works fine, but the openjdk build gives:

dyld: Symbol not found: __cg_jpeg_resync_to_restart Referenced from: /System/Library/Frameworks/ImageIO.framework/Versions/A/Resources/libTIFF.dylib Expected in: /Users/dicej/trash/openjdk-1.7.0-u80-unofficial-macosx-x86_64-image/jre/lib/libJPEG.dylib in /System/Library/Frameworks/ImageIO.framework/Versions/A/Resources/libTIFF.dylib

I've run into this before, so it isn't a new regression. Apparently the system libTIFF is trying to use the JRE's libJPEG and I'm not sure how to prevent that. Probably some dlopen magic, but I'm not ready to pursue it now, and maybe it's only a problem on my machine anyway.

Also, I can't find any pre-built OpenJDK 8 images, and building it from source requires downgrading to Xcode 4, which sounds like a nightmare. Someone who really cares about Avian/OpenJDK 8 on Mac can do that.

joshuawarner32 commented 8 years ago

Do you know where the try and except statements are used in the openjdk code? I'm worried that there's an interesting case that we're missing in our test suite where a windows structured exception will be thrown in such a block, not get caught, and lead to bizarre bugs down the road.

dicej commented 8 years ago

java_props_md.c

I do agree that my preprocessor hack is dangerous, so we could change it to a patch instead. Short of using a compiler that actually supports SEH, I don't know what a proper patch would look like.

joshuawarner32 commented 8 years ago

Technically, I think we could simulate try/except with the preprocessor and windows API calls (the same ones we use in windows/signal.cpp, IIRC), but that may be premature. I'd definitely be more comfortable with calling out each of the locations explicitly in a patch, since that'll give a more visible breadcrumb to investigate, should we ever run into troubles with this.

dicej commented 8 years ago

I've pushed a version that uses a patch instead of the preprocessor. It's even uglier, but safer.

joshuawarner32 commented 8 years ago

Looks like the Trace test failed again. It's gotten obnoxious enough that I'm tempted to do something like just disabling it on PR builds. What do you think?

dicej commented 8 years ago

Works for me. On Jan 13, 2016 12:17 PM, "Joshua Warner" notifications@github.com wrote:

Looks like the Trace test failed again. It's gotten obnoxious enough that I'm tempted to do something like just disabling it on PR builds. What do you think?

— Reply to this email directly or view it on GitHub https://github.com/ReadyTalk/avian/pull/472#issuecomment-171403045.