Rewritten to add content-type support.
* src/DownloadHandler.{h, cc}
* src/BtPostDownloadHandler.{h, cc}
* test/BtPostDownloadHandlerTest.cc
* src/MetalinkPostDownloadHandler.{h, cc}
* test/MetalinkPostDownloadHandlerTest.cc
* src/PostDownloadHandler.{h, cc}
* src/DownloadHandlerConstants.{h, cc}
* src/RequestGroup.cc
* src/HttpResponseCommand.cc
* src/FtpNegotiationCommand.cc
* src/SingleFileDownloadContext.{h, cc}
* src/RequestGroup.h
* src/RequestGroupCriteria.h
* src/ContentTypeRequestGroupCriteria.h
Added 'mem' option value for --follow-metalink,
--follow-torrent.
If it is give, metalink/torrent file is not written to the disk,
but
just is kept in memory. Parsing is occurred on memory.
* src/MetalinkHelper.{h, cc}
* src/MetalinkProcessor.h
* src/Xml2MetalinkProcessor.{h, cc}
* test/Xml2MetalinkProcessorTest.cc
* src/DownloadHandlerFactory.{h, cc}
* test/DownloadHandlerFactoryTest.cc
* src/PreDownloadHandler.{h, cc}
* src/OptionHandlerFactory.cc
* src/DefaultBtContext.{h, cc}
* test/DefaultBtContextTest.cc
* src/version_usage.cc
* src/Metalink2RequestGroup.{h, cc}
* src/RequestGroup.{h, cc}
* src/a2functional.h
* test/a2functionalTest.cc
* src/MemoryBufferPreDownloadHandler.{h, cc}
* src/OptionHandlerImpl.h
* src/prefs.h
* src/Util.{h, cc}
* test/UtilTest.cc
Keep DownloadResult rather than RequestGroup after downloads to
reduce
memory usage.
* src/RequestGroupMan.{h, cc}
* src/DownloadEngine.cc
* src/BtDependency.{h, cc}: Changed the type of dependee from
WeakHandle to SharedHandle because WeakHandle could be null.
* src/RequestGroup.{h, cc}
* src/DownloadEngineFactory.cc
* src/DownloadResult.h
Set totalLength after download finished
* src/UnknownLengthPieceStorage.{h, cc}
Keep torrent file specified in metalink in memory.
* src/Metalink2RequestGroup.cc
* src/BtDependency.cc
* src/TrueRequestGroupCriteria.h
Fixed the bug: seekg is used where seekp should be used.
* src/ByteArrayDiskWriter.cc
* test/ByteArraydiskWriterTest.cc
Reflect the download length of in-flight pieces.
It makes the download length readout more precise.
* src/DefaultPieceStorage.{h, cc}
* test/DefaultPieceStorageTest.cc
* src/a2functional.h
* test/a2functionalTest.cc
Lower CPU load when --max-download-limit is used.
There is up and down in speed indicator when enabling
http-pipelining but a download goes well. I think the problem is
that
because http-pipelining is enabled, DownloadCommand is created
for
each segment and in its constructor, PeerStat::downloadStart()
is
called. In PeerStat::downloadStart(), speed calculation object
is
reseted, which makes download speed zero.
* src/DownloadCommand.cc
Rewritten using accumulate.
* src/RequestGroupMan.cc (calculateStat)
Code clearnup.
* src/FtpNegotiationCommand.cc
* src/HttpResponseCommand.cc