I figured the problem, the typedef DexTypeItem was getting padded to 4 bytes even though it has only one u2 property. This was causing the structure array to be misaligned. The strange thing is when I compiled libdex, I passed the same gcc options as that of in the Android environment.
Now I am debugging the next issue "Too many exceptions during init". Any pointers? r...@172:/system/bin# dalvikvm -Xbootclasspath:/system/framework/ core.jar -classpath /data/HelloWorld.jar org.apache.HelloWorld V( 5387) VM init args (3): (dalvikvm) V( 5387) 0: '-Xbootclasspath:/system/framework/core.jar' (dalvikvm) V( 5387) 1: '-classpath' (dalvikvm) V( 5387) 2: '/data/HelloWorld.jar' (dalvikvm) V( 5387) VM options (3): (dalvikvm) V( 5387) 0: '-Xbootclasspath:/system/framework/core.jar' (dalvikvm) V( 5387) 1: '-classpath' (dalvikvm) V( 5387) 2: '/data/HelloWorld.jar' (dalvikvm) V( 5387) Using executionMode 2 (dalvikvm) V( 5387) mterp: interp is 16384 bytes, sisters are 2608 bytes (dalvikvm) V( 5387) Creating VM heap of size 2097152 (dalvikvm-heap) V( 5387) threadid=3: interp stack at 0x40016000 (dalvikvm) V( 5387) LinearAlloc: created region at 0x4153b000-0x4193afff (dalvikvm) V( 5387) Opening archive '/system/framework/core.jar' 0xbe877ab4 (dalvikvm) V( 5387) +++ numEntries=26 cdOffset=1557197 (dalvikvm) V( 5387) Couldn't open /system/framework/core.odex: No such file or directory (dalvikvm) V( 5387) Cache file for '/system/framework/core.jar' 'classes.dex' is '/data/dalvik-cache/sys...@framework@core....@classes.dex' (dalvikvm) V( 5387) dvmDexCacheStatus: Checking cache for /system/framework/ core.jar (/data/dalvik-cache/sys...@framework@core....@classes.dex) (dalvikvm) V( 5387) DexOpt: locking cache file /data/dalvik-cache/ sys...@framework@core....@classes.dex (fd=4, boot=1) (dalvikvm) V( 5387) DexOpt: locked cache file (dalvikvm) V( 5387) checking deps, expecting vfy=1 opt=1 (dalvikvm) W( 5387) DexOpt: incorrect opt magic number (0xff ff ff ff) (dalvikvm) D( 5387) Stale deps in cache file; removing and retrying (dalvikvm) V( 5387) DexOpt: locking cache file /data/dalvik-cache/ sys...@framework@core....@classes.dex (fd=4, boot=1) (dalvikvm) V( 5387) DexOpt: locked cache file (dalvikvm) V( 5387) DexOpt: successfully initialized new cache file (dalvikvm) D( 5387) DexOpt: --- BEGIN 'core.jar' (bootstrap=1) --- (dalvikvm) V( 5387) DexOpt: waiting for verify+opt, pid=5388 (dalvikvm) V( 5388) Creating VM heap of size 2097152 (dalvikvm-heap) V( 5388) threadid=3: interp stack at 0x40016000 (dalvikvm) V( 5388) LinearAlloc: created region at 0x4153b000-0x4193afff (dalvikvm) V( 5388) Continuing optimization (/system/framework/core.jar, isb=1, vfy=1, opt=1) (dalvikvm) V( 5388) +++ swapping and verifying (dalvikvm) V( 5396) +++ DEX 0x191e0: allocateAux 27166+2910+24694+9879 * 4 = 258596 bytes (dalvikvm) V( 5396) Class lookup: classes=2760 slots=8192 (33% occ) alloc=98312 total=768 max=10 (dalvikvm) V( 5396) DexOpt: +++ trying to load 2760 classes (dalvikvm) V( 5396) +++ threadid=3: added 0x70cd0, now 1 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/Class;' (nil) (isnew=1) --> 0x404fb340 (dalvikvm) V( 5396) +++ threadid=3: added 0x70ce8, now 2 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/Object;' (nil) (isnew=1) -- > 0x404fb3e8 (dalvikvm) V( 5396) +++ threadid=3: removed 0x70ce8, now 1 (dalvikvm) V( 5396) +++ threadid=3: added 0x70ce8, now 2 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/io/Serializable;' (nil) (isnew=1) --> 0x404fb490 (dalvikvm) V( 5396) +++ threadid=3: removed 0x70ce8, now 1 (dalvikvm) V( 5396) +++ threadid=3: added 0x70ce8, now 2 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/reflect/ GenericDeclaration;' (nil) (isnew=1) --> 0x404fb538 (dalvikvm) V( 5396) +++ threadid=3: removed 0x70ce8, now 1 (dalvikvm) V( 5396) THROW 'Ljava/lang/NoClassDefFoundError;' msg='D' cause= (none) (dalvikvm) V( 5396) +++ threadid=3: added 0x70ce8, now 2 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/ NoClassDefFoundError;' (nil) (isnew=1) --> 0x404fb5e0 (dalvikvm) V( 5396) +++ threadid=3: added 0x70d40, now 3 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/LinkageError;' (nil) (isnew=1) --> 0x404fb688 (dalvikvm) V( 5396) +++ threadid=3: added 0x70d78, now 4 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/Error;' (nil) (isnew=1) --> 0x404fb730 (dalvikvm) V( 5396) +++ threadid=3: added 0x70db0, now 5 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/Throwable;' (nil) (isnew=1) --> 0x404fb7d8 (dalvikvm) V( 5396) --- LinearRealloc(84) old=92 (dalvikvm) V( 5396) +++ threadid=3: removed 0x70db0, now 4 (dalvikvm) V( 5396) +++ threadid=3: removed 0x70d78, now 3 (dalvikvm) V( 5396) +++ threadid=3: removed 0x70d40, now 2 (dalvikvm) V( 5396) +++ threadid=3: removed 0x70ce8, now 1 (dalvikvm) V( 5396) +++ threadid=3: added 0x70ce8, now 2 (dalvikvm) V( 5396) +++ threadid=3: added 0x70d78, now 3 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/String;' (nil) (isnew=1) -- > 0x404fb880 (dalvikvm) V( 5396) +++ threadid=3: added 0x70db0, now 4 (dalvikvm) V( 5396) +++ dvmAddClassToHash 'Ljava/lang/CharSequence;' (nil) (isnew=1) --> 0x404fb928 (dalvikvm) V( 5396) +++ threadid=3: removed 0x70db0, now 3 (dalvikvm) V( 5396) THROW 'Ljava/lang/NoClassDefFoundError;' msg='F' cause= (none) (dalvikvm) E( 5396) Too many exceptions during init (failed on 'Ljava/lang/ NoClassDefFoundError;' 'F') (dalvikvm) E( 5396) VM aborting (dalvikvm) W( 5395) DexOpt: --- END 'core.jar' --- status=0x008b, process failed (dalvikvm) E( 5395) Unable to extract+optimize DEX from '/system/framework/ core.jar' (dalvikvm) V( 5395) Closing archive 0xbe825aa4 (dalvikvm) V( 5395) munmap(0x4193b000, 1559732) succeeded (dalvikvm) D( 5395) Failed on '/system/framework/core.jar' (boot=1) (dalvikvm) V( 5395) VM shutting down (dalvikvm) D( 5395) VM cleaning up (dalvikvm) V( 5395) Unmapping linear allocator base=0x4153b000 (dalvikvm) D( 5395) LinearAlloc (nil) used 4100 of 4194304 (0%) (dalvikvm) W( 5395) JNI_CreateJavaVM failed (dalvikvm) Dalvik VM init failed (check log file) On Mar 12, 7:54 pm, fadden <fad...@android.com> wrote: > On Mar 12, 10:02 am, Sean McNeil <seanmcne...@gmail.com> wrote: > > > I'm not sure, but could this also be related to short enums? I seem to > > remember seeing something like this when I had a mixture of old code > > compiled with and new code compiled without short enums. > > Everything related to DEX file access should be using the "physical" > size types; by tradition u1/u2/u4/u8. Optimized DEX files are allowed > to be non-portable, but the base classes.dex should not be subject to > the whims of compilers (either while reading or writing). > > Of course, if half the code is compiled one way and half the other > way, many obscure failure modes are possible. > > The easiest way to debug this might be to run the VM on the same > core.jar on a working platform, and step through both simultaneously > (either with a debugger or with logs) and identify the point where one > goes off the rails. --~--~---------~--~----~------------~-------~--~----~ unsubscribe: android-porting+unsubscr...@googlegroups.com website: http://groups.google.com/group/android-porting -~----------~----~----~----~------~----~------~--~---
