Releases: lingochamp/FileDownloader
Releases · lingochamp/FileDownloader
v1.7.5
Version 1.7.5
2018-08-03
Fix
- Fix: fix raise "Not allowed to start service Intent..." issue when starting DownloadService on Android O and the application isn't on the foreground and also not on the whitelist, because we can't use
JobSchedulerto handle the download affair. closes #1078
Enhance
v1.7.4
Version 1.7.4
2018-05-19
Fix
v1.7.3
v1.7.2
Version 1.7.2
2018-03-14
Fix
- Fix: do not download callback error when the instance length of the task is zero, callback complete directly instead. closes #789
- Fix: fix the temp duplicate data in the database isn't removed when there is another running task with the same temp file path. closes #953
- Fix: the data lost when retry. closes #949
- Fix: fix the instance-length is always
1when theContent-Rangeisn't provided but theContent-Lengthis provided on the trial connection.
Enhancement
- Improve Practicability: using the content length value on the
Content-Rangefield when there isn'tContent-Lengthfield found in the response header. closes #967
v1.7.1
v1.7.0
Fix
- Fix: fix update status can't keep flow through making updating status synchronized with pause action. close #889
- Fix: fix the sofar-bytes carry back through pending state callback has already discarded. close #884
- Fix: fix can't find filename if filename value on content-disposition without around with ". close #908
- Fix: correct
setCallbackProgressTimesmethod makesetCallbackProgressTimeswork correctly. close #901 - Fix: fix download useless data on tcp-window because of the first trial connection use
0-infiniterange. close #933 - Fix: close input stream when connection ending avoids input-stream leak especially for the trial connection.
Enhancement
- Improve Practicability: do not remove the temp-file if rename it to the target path success to prevent raise some file-system problem. close #912
- Improve Practicability: discard range totally if the range is right but backend response 416. close #921
- Improve Performance: using HEAD request method instead of GET method for trial connect. ref #933
v1.6.9
Version 1.6.9
2017_12_16
Fix
- Fix(serial-queue): fix deadlock on
FileDownloadSerialQueue. closes #858 - Fix: do not use j-unit on library non-unit-test env to fix the
no-static-method-foundissue on some mi-phones. closes #867 - Fix: fix decrease two times of retry-chance each time of retry. closes #838
- Fix: fix get status is pending when a task has been paused on pending status. closes #855
Enhancement
- Improve Practicability: public
SqliteDatabaseImpl、RemitDatabase、NoDatabaseImpl, so you can overwrite them - Improve Practicability: support downgrade version from newer version
- Improve Practicability: add the default
User-Agentif upper layer does not add. closes #848 - Improve Performance: change the keepalive second(5s to 15s) for each executor, since when downloading multiple tasks thread release and recreate too frequently
- Improve Performance: using
RemitDatabaseinstead ofDefaultFiledownloadDatabaseto avoid some small task start and finished on the very short time but consume too much time on sync status to database what is useless
v1.6.8
- Fix: fix resume from breakpoint failed because of
isAlivenot guarantee on Network-thread. this closes #793 - Fix: fix resume from breakpoint failed, because of multi-thread update status very frequently and Messenger can't guarantee order. this refs #793, #764, #721, #769, #763, #761, #716
- Fix: do not crash user when a task has finished but the messenger still has messages, because it's fine for the user. this closes #562
- Fix: fix the callback error of 'it can't take a snapshot for the task xxx' when a user invokes pause very frequently.
- Fix: fix the case of process on the model is wrong which raise 416 each time when restarting it.
v1.6.7
Fix
- Fix: Avoid error/pause status is covered by other processing-status which will cause resume-failed, task-never-end. this closes #769, closes #764, closes #761, closes #763, closes #721, closes #716
- Fix: Fix request range value turn to negative when resuming a task which has a process more than 1.99G on its one block. Thanks to @hongbiangoal closes #791
v1.6.6
Fix
- Fix(file-integrality): update the process to the database only if all buffers on output-stream or system have been flush and sync to device successfully to avoid resume on the wrong point raise complete file, not integrality. Closes #745
- Fix(clear): fix
FileDownloader#clearAllTaskDatanot clear connection table. Closes #754
Enhancement
- Import Performance: optimize the default output-stream with buffered-output-stream, now the VM buffers length is 8192 bytes.
