| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
commit 3b43a223253176731567c6d8b7a67c9a6110782b
|
|
|
|
| |
'os.and.arch', instead of android.abi for harmonization
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
matching StaticConfig at runtime; Fix PPC (Bug 1056) and MIPSLE (Bug 1014) issues.
Currently the StaticConfig is being queried
via the key[OSType, CPUType ..]
as pre-determined by Java properties or the ELF parser.
This adds complication to maintain different platforms
and the key query might not even be sufficient.
The MachineDescriptor's StaticConfig only purpose shall be
to speed-up native data size and offset/alignment retrieval.
This is done by using the StaticConfig index within
all StaticConfig[]s as a lookup-index for the precomputed
struct's size and offset tables.
+++
Solution:
Rename: MachineDescriptor -> MachineDataInfo
Rename: MachineDescriptorRuntime -> MachineDataInfoRuntime
After having defined os.and.arch (OSType, CPUType and ABIType)
w/ the optional help of the now self containing ELF Reader (Bug 1125),
the native gluegen-rt library gets loaded enabling JNI methods.
It is satisfactory to retrieve MachineDataInfo
at runtime w/ JNI and find the matching/compatible StaticConfig.
Only in case none is found, the program needs to abort.
Otherwise the found MachineDataInfo.StaticConfig and MachineDataInfo
are stored for further use (see above).
This removes above complication and key to StaticConfig mapping.
New platforms simply need to add a new unique entry into the
StaticConfig[] table.
++
Also fixes Bug 1056 (PPC), thanks to tmancill [@] debian [.] org,
and Bug 1014 (MIPSLE), thanks to Dejan Latinovic.
Parts of the patch for Bug 1014 from Dejan Latinovic are included.
also solved by this change set.
|
|
|
|
| |
detection
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
ELF Reader 'jogamp.common.os.elf' currently uses
Platform's pre-determined OS_TYPE and CPUType.
It also uses the host platforms MachineDescription,
hence can not read ELF files from other machines.
This also forbids Platform to determine CPUType etc
w/o having a valid 'os.arch' property.
+++
ElfHeader should be split in
- ElfHeaderPart1 (CPUType independent)
- ElfHeaderPart2 (CPUType dependent)
Fix shall make the ELF Reader self containing
by only using ELF CPUType data, etc.
This requires customization of struct parsing,
where MachineDescription.Static index shall be
- defined in ElfHeaderPart1 using e_Ident's CPUType.
- used in ElfHeaderPart2 and all its struct types.
|
|
|
|
|
|
|
|
|
|
|
|
| |
Android:
- Detect ABIType.EABI_GNU_ARMHF via 'armeabi-v7a-hard'
Platform.CPUType:
- contains is32Bit now
MachineDescription:
- Rename *x86_64_unix* -> *lp64_unix*, reflecting universal __LP64__ mode
- Remove is32Bit, which is determined by CPUType
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
and fix Android AArch64 BionicDynamicLinker (Bug 1122)
- Bulk Permissions
ProcAddressTable.reset(..) performs address lookup in one block.
Now claiming all permissions upfront once, and releasing them afterwards.
- Cleanup DynamicLinker impl.
Proper top-down impl. of DynamicLinkerImpl,
handling all security code and validations.
- Fix Android AArch64 BionicDynamicLinker (Bug 1122)
Dalvik uses diff RTLD_* defines for AArch64!
|
|
|
|
| |
Android compilerflags
|
|
|
|
|
| |
Symbol could not be found w/ Android:
UnixDynamicLinkerImpl_JNI.c::Java_jogamp_common_os_UnixDynamicLinkerImpl_dlerror
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Add AArch64 detection via
- Elf Parser
- Android properties
- Java properties
- Android: Validate CPUType.Family _and_ ABIType
- MachineDescription
- Remove redundant Type ID and its field
- Reuse X86_64_UNIX for AArch64 (static config)
New ARCH 'aarch64' for
ant: <os arch>
armv8a
aarch64
New CPUType.ARM64 (ARM):
java: os.arch
aarch64
arm64
New CPUType.ARMv8_A (ARM):
java: os.arch
armv8-a
arm64-v8a
New ABIType:
EABI_AARCH64
|
|
|
|
| |
Shdr and Ehdr via GlueGen
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
(lollipop/art)
Our Android ClassLoaderUtil, used to daisy chain dex ClassLoader
using GlueGen, Joal and Jogl APK files, doesn't work
on Android 5.* (lollipop) with the art runtime.
See below adb log.
Seems like this is a known issue
https://code.google.com/p/android/issues/detail?id=82122
Workaround mentioned is to copy the APK file(s)
to a temp directory ..
|
|
|
|
|
|
|
| |
>= 10d
Adjust build scripts and GlueGen Android tasks
for SDK >= 24.0.2 (build-tools >= 21.1.2) and NDK >= 10d
|
| |
|
|
|
|
| |
'GetProcAddressTableExpr'
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Windows via BATCH file execution
Batch file execution test via direct call, i.e.'a.bat',
does not work on some Windows machine, not reproducible here!
A workaround would be to call the batch file explicit via 'CMD /c a.bat':
- works when using 'Software Restriction Policies' (Bug 1015 Comment 2)
- does _not_ work when denying ACL "Traverse Folder / Execute File" permission (Bug 1015 Comment 3)
Due to this bug, we need to use a native execution:
- Performing executable test w/ native exe file instead of batch file
on Windows x86 32bit and 64bit,
- using [1] TinyPE XP-W8 Compatible - x86 32bit and 64bit - 268 bytes
- Tested on: WinXP-32bit, Win7-64bit and Win8-64bit
- Author/License: Ange Albertini, BSD Licence, 2010-2013
- On all other Windows machines (ARM, ..),
we still use direct execution of 'a.bat'
but may add native exe files for missing platforms.
+++
This patch injects said binaries within the java jar file
and copies it into the 'to be tested' temp folder for execution.
[1] TinyPE XP-W8 Compatible - x86 32bit and 64bit - 268 bytes
is included within 'src/native/tinype-corkami',
build manually and imported as 'src/java/com/jogamp/common/util/bin/exe-windows-i586-268b.bin'.
+++
|
| |
|
| |
|
|
|
|
|
|
|
| |
incl. ProcAddressTable
Emitted Java classes as well as optional ProcAddressTable shall
use unique generated access modifiers covering all MethodAccess cases.
|
|
|
|
|
|
|
|
|
|
|
|
| |
access permission on Windows via BATCH file execution
Try using explicit call to Windows 'cmd.exe' w/ referencing the BATCH file location,
via
'Runtime.getRuntime().exec(new String[] { "cmd", "/c", "a.bat" } );'
While the bug itself could not be reproduced here,
I could test on Windows 7 (64bit and 32bit), as well as an WindowsXP 32bit
that no regression occured.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
setting of stream position (optional)
- Add setting position entry, optionally supported,
e.g. ByteBufferStream and ByteArrayStream
- Remove 'msbFirst' parameter on all 'bulk' read/write
operations.
These methods use LSB-first always, allowing proper
stream access of data w/ different bit-sizes.
Data is now read/write as little-endian
and swapped accordingly.
Optimizations are adopted for LSB-first operations.
This change removes API confusion/bugs:
- removes one decision (parameter)
- removes the data reversion case
- removes bugs w/ different bit-sizes
|
| |
|
| |
|
| |
|
|
|
|
| |
(500 ms) per JVM after test class (cooperative w/ other JVM test instances)
|
|
|
|
|
|
| |
- Increase poll frequency from every 1000 ms -> 500 ms
- After releasing the lock, sleep to 6 x poll period, i.e. 3s
|
|
|
|
| |
static fields; Change name to avoid testing itself.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
from JunitTracer and use it for all tests.
- Certain performance tests don't make sense utilizing several jenkins unit tests on one machine,
hence singleton test execution is desired.
- JOGL's UITestCase also will derive from this SingletonTestCase
|
|
|
|
|
|
|
|
|
|
|
| |
FLUSH_PRE_SOFT)
FLUSH_PRE_SOFT cannot be handled by some platforms, e.g. Windows 32bit.
FLUSH_PRE_HARD is the most reliable caching mode
and it will fallback to FLUSH_PRE_SOFT if no method for 'cleaner' exists.
Further, FLUSH_PRE_HARD turns our to be the fastest mode as well.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Validate active and GC'ed mapped-buffer count
in cleanAllSlices() via close() ..
- Fix missing unmapping last buffer in notifyLengthChangeImpl(),
branch criteria was off by one.
- cleanSlice(..) now also issues cleanBuffer(..) on the GC'ed entry,
hence if WeakReference is still alive, enforce it's release.
- cleanBuffer(..) reverts FLUSH_PRE_HARD -> FLUSH_PRE_SOFT
in case of an error.
- flush() -> flush(boolean metaData) to expose FileChannel.force(metaData).
- Add synchronous mode, flushing/syncing the mapped buffers when
in READ_WRITE mapping mode and issue FileChannel.force() if not READ_ONLY.
Above is implemented via flush()/flushImpl(..) for buffers and FileChannel,
as well as in syncSlice(..) for buffers only.
flush*()/syncSlice() is covered by:
- setLength()
- notifyLengthChange*(..)
- nextSlice()
Always issue flushImpl() in close().
- Windows: Clean all buffers in setLength(),
otherwise Windows will report:
- Windows: Catch MappedByteBuffer.force() IOException
- Optimization of position(..)
position(..) is now standalone to allow issuing flushSlice(..)
before gathering the new mapped buffer.
This shall avoid one extra cache miss.
Hence rename positionImpl(..) -> position2(..).
- All MappedByteBufferOutputStream.write(..) methods
issue syncSlice(..) on the last written current slice
to ensure new 'synchronous' mode is honored.
+++
Unit tests:
- Ensure test files are being deleted
- TestByteBufferCopyStream: Reduced test file size to more sensible values.
-
|
|
|
|
| |
to stream copy as well as direct memory mapped ByteBuffer access
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
OutputStream impl.
Added MappedByteBufferOutputStream as a child instance of MappedByteBufferInputStream,
since the latter already manages the file's mapped buffer slices.
Current design is:
- MappedByteBufferInputStream (parent)
- MappedByteBufferOutputStream
this is due to InputStream and OutputStream not being interfaces,
but most functionality is provided in one class.
We could redesign both as follows:
- MappedByteBufferIOStream (parent)
- MappedByteBufferInputStream
- MappedByteBufferOutputStream
This might visualize things better .. dunno whether its worth the
extra redirection.
+++
MappedByteBufferInputStream:
- Adding [file] resize support via custom FileResizeOp
- All construction happens via ctors
- Handle refCount, incr. by ctor and getOutputStream(..), decr by close
- Check whether stream is closed already -> IOException
- Simplify / Reuse code
MappedByteBufferOutputStream:
- Adding simple write operations
|
|
|
|
| |
IOException (Add note to FLUSH_NONE); Reduce test load / duration.
|
|
|
|
| |
getTempDir(..) throws IOException
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
InputStream impl., incl. mark/reset
- ByteBufferInputStream simply impl. InputStream for an arbitrary 2MiB restricted ByteBuffer
- Users may only need a smaller implementation for 'smaller' file sizes
or for streaming a [native] ByteBuffer.
- MappedByteBufferInputStream impl. InputStream for any file size,
while slicing the total size to memory mapped buffers via the given FileChannel.
The latter are mapped lazily and diff. flush/cache methods are supported
to ease virtual memory usage.
- TestByteBufferInputStream: Basic unit test for basic functionality and perf. stats.
|
|
|
|
|
|
|
|
|
|
| |
RuntimeException for IO failures
- Fix API doc
- copy*() methods only close stream they open, no need to mention the proper behavior (obvious).
- getTempDir(..) shall throw IOException instead of RuntimeException for IO failures
- Ease using getTempDir(..), i.e. only handle IOException
|
|
|
|
|
|
|
| |
9cdb6bbf6d3248a65aa78ced08211d609c356f91
commit 9cdb6bbf6d3248a65aa78ced08211d609c356f91 fixes the tests of the Class, Field and Method accessor bit
for compatibility.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
exetst.getCanonicalPath() command
Having a test-executable file with SPACE in it's temporary path,
e.g.
Unix: /magic temp/user/
Windows XP: C:\Documents and Settings\user\temp
results to split up commands if using
'Process.exec(String)', since it utilizes
a StringTokenizer to assemble the command String[].
This patch uses 'Process.exec(String[])' avoiding
splitting up SPACE in the full command string/path.
Major impact is on 'Windows XP',
where a SPACE separated TEMP/TMP folder is the default.
Note-1: Commit 9bc3d3f78bb2fb1aa0ccfb02ffb5bdda74420cac
introduced testing executable permission on Windows (Bug 1015)
and rendered this issue visible.
To reproduce the issue on Windows:
+++
set TEMP=C:\Documents and Settings\jogamp\temp
set TMP=C:\Documents and Settings\jogamp\temp
"C:\Program Files (x86)\Mozilla Firefox\firefox.exe"
+++
Note-2: This bug affects all platforms!
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
for getNormalized(), getDirectory(), getParent() and getRelativeOf()
- Refine API doc
- Add notion of {@code host} and {@code port} validation
- Add create(Encoded ..), allowing creation of variants w/o re-encoding
- Provide common impl. for getNormalized(), getDirectory(), getParent() and getRelativeOf()
Above feature methods share common goals, hence use same implementation:
- If opaque, cut-off query and merge after operation
- cleanup path, i.e. /dummy/../test/ -> /test/
- cutoff file, dir - if requested
- append optional appendix and cleanup again
Return behavior various thought, i.e. null, this or allow exception.
Enhanced test of above features.
|
| |
|
| |
|
|\ |
|