Gnutella Forums - 2.8.10.98 (at least) issue

Gnutella Forums (https://www.gnutellaforums.com/)

- Feature Requests & Bug Reports (https://www.gnutellaforums.com/feature-requests-bug-reports/)

- - 2.8.10.98 (at least) issue (https://www.gnutellaforums.com/feature-requests-bug-reports/70894-2-8-10-98-least-issue.html)

2.8.10.98 (at least) issue

I use Mac OS X.3.9, so am limited to Phex 2.8.10.98 so this issue may not be present in newer versions. I seem to get "crashes" after extended sessions (maybe several days) signaled by a periodic "too many open files" and some other alerts that come and go from the right lower corner of the screen. I say "crashes" because Phex no longers seems to be responsive to controls or, sometimes, it's responsive but will not quit despite running through the visual part of the process (repeatedly if one cares to do so). Thus I have to force quit.

The problem comes in when the force quit occurs in such a way that Phex "forgets" what it was doing. IOW, the .phex folder contents gets crumped so that what Phex was doing gets lost (doesn't always happen - probably a function of exactly what error got Phex into whatever limiting condition it's reporting). This can leave a number of INCOMPLETE files in the Incomplete directory which no longer are known to Phex when restarted.

Perhaps there's some means, as there is in Limewire via its Library view, of continuing previously aborted downloads. If so, I've been unable to locate it.

A real downside of this behavior is that incomplete files build up in a place that users don't typically look, gobbling up space.

Yes... various of the problems you described are known in 2.8..

I hope the answer I gave here in the other thread is explaining our problem...
http://www.gnutellaforums.com/featur...nst-3-0-x.html

In a generic way, I suppose it does. Can you point me to specific data about why 1.5 and 1.6 are so very advantageous that using them is worth cutting off large numbers of potential users? I've been programming for decades but don't Java jive.

A quick note to your original problem since I just looked through 2.8.10 code.

Your problems mainly appear in two places, that is:
phex/host/CaughtHostsContainer.java
diff is here:
SourceForge.net Repository - [phex] Diff of /phex/tags/release-3.0.2.100/src/phex/host/CaughtHostsContainer.java

and
phex/download/swarming/SWDownloadFile.java
phex/common/file/ManagedFile.java
described in the change set:
SourceForge.net Repository - [phex] Revision 3569

It should be possible to backport the stuff to Java 1.4 when you just ignore the Java 5 specific changes.

Now all I have to do is figure out how to check out the 2.8.10.98 code with svn and get java stuff compiling on this rig. Maybe someday I'll know my way around the program enough to help out. Don't hold your breath. ;-)

compiling 2.8.10.98 on Mac OS X.3

FWIW, I managed to checkout the 2.8.10.98 version via svn (thanks - svn is new to me as well as java). I have to say that, for Mac OS X at least, the "directions" leave something to be desired, for a clueless newbie anyway. After reading things, I found that I needed "ant". A version of ant comes with Tiger (X.4.?) but apparently not with Panther.

Getting ant compiled and installed is also a process fraught with misdirections to avoid. Panther doesn't seem to like ant 1.7 (the JUnit issue, if nothing else). I have no idea whether I've got it right or not but some manner of compilation of Phex 2.8.10.98 took place. I've yet to check out whether it's what I have in the packaged app I'm using or not. It compiled more quickly than I was expecting it too. Makes me wonder if I'm not doing something really wrong.

I think Arne, if he is around, can tell you more about ant in OS X. But AFAIK the ant binary distribution, which you should be able to use without problems, contains a shell script to execute ant easily.
For Phex it should be fine to use ant 1.6.5, not sure if its easier, but I also still use it.
Phex compilation doesn't take very long. If you have a output\lib\phex.jar file you can expect that the build finished correctly.