#1
|
|||
|
|||
removal of files on selecting "remove from list"
jdownloader version: 0.9.579
Java version: sun-java6 (6-20-0-lenny1) OS: debian lenny no FW / AV Software following situation: I have a list of 1 file with 8 mirrors. one mirror starts the download, while the download runs I mark all other mirrors and select "remove -> from list" (NOT "from list and disc"!) after clicking I can see the .part-file disappearing in download-directory. the download is still running. with lsof I can see the file listed to the according PID, but with "deleted" afterwards. after the download is finished the file is deleted complete. I hope this is enough, should be simply reproduceable. best regards Last edited by Cyber; 10.10.2010 at 21:59. |
#2
|
||||
|
||||
Please check whether this still occurs in the Nightly test release.
Instead of using "Remove - from list", please just disable the mirrors until the file has been downloaded and checked. This will save you other headaches. |
#3
|
|||
|
|||
actually I did not have the possibility to test this with the nightly build, but I just took a look with the "normal" branc, full updated, it still had the issue.
but I think this is normal as you told me I have to try with a nightly build ... btw.: the reason why I delete them out of the list: jdownloader consums a lot more system ressources if the list is quite big (> 50 items) - as my system is very small I try to reduce the list as soon as possible ... |
#4
|
||||
|
||||
I understand and agree with your concerns. I currently have trouble keeping track of what I have downloaded and where it is. I have thousands of links I am interested in downloading stored in DLCs, so they do not have to be in memory.
JD stores the entire database in memory and in an inefficient way. This is expected to change in the future. One feature request is to store completed and disabled links separately from links that are downloading or waiting to download. This will allow users to keep all completed links and packages around. Another feature request is to add tags to packages, so they are easier to find. __________________ I understand the limitations of small systems. I use a 9 year old computer (see note 1). But, for packages that contains only one archive, the extra links for disabled mirrors are only a small increase. I currently have almost 3000 links and JD is running with less than 175MiB. If I run JD with no links, it takes around 90MIB. This is less than 32KiB per link (this is more than it should take, but extensive database work is planned for the future). However, if you have 40 links (two large archives) plus 80 mirrors (two mirrors each), that is only 120 links total. The total for the links is less than 4 MiB of memory. Even on a computer with only 256 MiB of memory this is not a hardship unless you are overloading the system in some other way. As for CPU time, extra links do not take extra download time if the JD window is minimized. It is the GUI and database that slow down with extra links. _____________________ This assumes you are using low impact security software (as I recommend in the Self Help Index). If you are using an integrated security package (such as KIS or ESET), up to 1/3 of your CPU time can be expended by the security package. A minimum security configuration for Windows (Microsoft Security Essentials with no firewall) is not noticeable. ________________________ Overall, the nightly test build is around twice as fast as the Stable (normal) branch. This represents the kind of improvement we can expect in the next release. The nightly build is not intended for daily use, but for testing. I break that rule, but it is with the knowledge of the support staff and I understand that I am risking a lot by doing this. So far, the worst that has happened is that my links were read wrong from the library and all were disabled (I enabled them and everything was fine). Some features disappear occasionally and the UI changes can be surprising. drbits Note 1: I used to write programs for computers with 256KB of memory and 20MB of disk space. I used to comb through the code looking for things to shrink. The documents we were dealing with were often larger than all of the memory on the computer and the program would not all fit in memory either. I had to manually control which parts of the documents were in memory and which parts of the program were in memory. On another project, the computer was just fast enough to read its input and copy it to memory. We looked at each low level instruction and the amount of time each instruction took. We got the loop down to 8 instructions. Note 2: MiB = 2 ^ 20 this is a relatively new convention. MB can mean the same as MiB or it can mean 1 000 000. Disk drives are measured in GB or TB (10^9 or 10^12), which is smaller than the expected GiB 2^30 or TiB 2^40. (a 1 TB drive holds around 930 GiB, about 10% less that 1 TiB). It is easiest to see the difference with KB (1000) versus KiB (1024). There is also a convention that Kb is 1024 bits or 128 bytes. Note 3: For testing purposes, I have had over 10 000 packages and 70 000 links in JDownloader at one time. JDownloader worked fine, but the GUI was much too slow. I never recommend more than 1000 packages and 10 000 links at one time (the program has an acceptable response time at that level, provided you have not overloaded the computer with other programs). Last edited by drbits; 27.10.2010 at 05:17. |
#5
|
|||
|
|||
drbits thanks for the tip regarding minimizing jdownloader - my problem is solved, jdownloader only eats CPU if it is not minimized. as soon as it is minimized it just downloads ... :D
memory usage is not as important as cpu usage for me. my system is equipped with enough memory, but VIA-cpus are soooo slow ... >>>cyber |
Thread Tools | |
Display Modes | |
|
|