RssBandit / RssBandit

Browsing web news without a browser
https://www.rssbandit.org
Other
98 stars 24 forks source link

high CPU utilization after upgraded to 1.5.0.5 (Jubilee) #930

Closed clairernovotny closed 7 years ago

clairernovotny commented 17 years ago

Hi,

After upgraded to 1.5.0.5 from 1.3, RSSbandit occupies 100% CPU time. My OS is WinXP Pro, Eng edition, with all patches available on windowsupdate. Let me know if you need other info or log. My email address is pirate585@gmail.com.

Regards, Eric

Reported by: *anonymous

Original Ticket: "rssbandit/bugs/928":https://sourceforge.net/p/rssbandit/bugs/928

clairernovotny commented 17 years ago

Logged In: NO

it does that for me too sometimes, when it seems to get into a "Download all RSS entry images" in one go. I don't see why it tries to do that? Perhaps its try to pre-load/cache all images for me?

Original comment by: nobody

clairernovotny commented 17 years ago

Logged In: NO

I have the same issue.

I attached the debugger and it seems that something is causing an endless loop in the "Lucene.Net.Analyses.CJK.CJKTokenizer.Next() call.

Call stack: Lucene.Net.Analyses.CJK.CJKTokenizer.Next() Lucene.Net.Analysis.StopFilter.Next() Lucene.Net.Index.DocumentWriter.InvertDocument(...) Lucene.Net.Index.DocumentWriter.AddDocument(...) NewsComponents.Search.LuceneIndexModifier.AddSingleDocument(...) NewsComponents.Search.LuceneIndexModifier.PerformOperation(...) NewsComponents.Search.LuceneIndexModifier.ThreadRun()

Tried to set a break point on all of the methods on the callstack but they never got called and CPU stayed pegged to 100% for an hour.

Original comment by: nobody

clairernovotny commented 17 years ago

Logged In: YES user_id=714452 Originator: NO

Can you please provide the UI language you use? The index analyzer depends on it. This would help to track down to the causing lang. specific analyzer.

Original comment by: netsrotr

clairernovotny commented 17 years ago

Logged In: YES user_id=714452 Originator: NO

The CJK Analyzer/Tokenizer is used for japanese ("ja") and korean ("ko") languages. So if a weblog post item states it is on one of these languages, it comes into play to analyze/tokenize the post content for index. Can you please verify you subscribed to such feed(s) that provide japanese/korean posts? Can you provide the link(s) here?

Also: I attached a version of the lucene.net.dll for you to copy over the installation: does the CPU peg still happens? I removed one stopword I guessed causing the error: an empty one was listed (""). Please report back any issues and also the success case! File Added: Lucene.Net.1.9.1.2.CJK.stopwords.issue.zip

Original comment by: netsrotr

clairernovotny commented 17 years ago

slightly modified lucene.net.dll

Original comment by: netsrotr

clairernovotny commented 17 years ago

Logged In: YES user_id=714452 Originator: NO

Please see my post at http://www.rendelmann.info/blog/PermaLink.aspx?guid=87c2ab4e-b22c-4215-bc00-d6a2566d46e9 and test it. Do NOT use the attached DLL (to this bug report below), it does NOT fix the issue! Try the one mentioned and linked in my blog post!

Original comment by: netsrotr

clairernovotny commented 17 years ago

Logged In: YES user_id=714452 Originator: NO

OK, the new lucene.net.dll fix the problem

Original comment by: netsrotr