- From: Henrik Frystyk Nielsen <frystyk@w3.org>
- Date: Mon, 09 Mar 1998 18:49:04 -0500
- To: "QingLong" <qinglong@Yggdrasil.com>, www-lib@w3.org
At 22:58 3/9/98, QingLong wrote: > > Dear W3C WWW Reference Library Developers, Testers, Users. > > The attached is gzip'ed patch against w3c-libwww-5.1i that cures (I hope) > some of the library problems: Great - Thanks, I will see how it can be merged in with the 5.1j release. How is the relative performance when using pipelining in a GUI browser - does it make a difference? I would _really_ like to see that! > HTHost.c > ~~~~~~~~ > HTHost_decreaseRetry() > The bug is obviously a fatal typo. > It has already been pointed out here by > John Nunneley <johnn@dascom.com> more than half a year ago, > but the bug is still there. :( ok > HTHost_addNet() > I have successfully used libwww-5.0a with non-blocking sockets > (in the context of Arena browser), but failed with 5.1i. > The library began looping on inline documents (images). > Been investigated it turned out that it was spawning multiple > copies of the same net objects and adding them to `pending' list, > while the `pipeline' was occupied by the net object associated with > the main document, that actually has already been loaded > (but it's net object has not been deleted from pipeline). > I don't know what does cause that `pipeline jam', > is it error in libwww or in the application (Arena browser). > Anyway (I guess) the same net objects should not be added > mutiple times to either queue (`pending' or `pipeline' list), > so add check to the function whether the object is already > on either of the lists. > With the patch the browser appears to work well. Normally, libwww doesn't care about what the application requests but it bothers me that you are saying that it's the _same_ Net objects. I will have a closer look tonight. > > HTWAIS.c > ~~~~~~~~ I had more or less dumped WAIS - it it important for you to have? I can certainly take your patches and add them. > HTEscape.c > ~~~~~~~~~~ > from_hex() > The WAIS module needs this function, so make it public. ok. > HTList.c > ~~~~~~~~ > HTList_removeObjectAll() > The original function is unusable: > while object removal it first frees HTList element, > then tries to use it's contents (in `while(me->next)') causing, > of course, segmentation violation. Yes, that doens't look too good! -- Henrik Frystyk Nielsen, World Wide Web Consortium http://www.w3.org/People/Frystyk
Received on Monday, 9 March 1998 18:49:37 UTC