Version: (using Devel) Compiler: debian gcc 4.3.1 OS: Linux Installed from: Compiled sources Heres a screenshot of konqueror, and a bunch of broken images (mainly pngs): http://strangesoft.net/konquerorpngbug.png It used to work fine a week or three ago. definitely works in firefox (any version) or kde 3 (any version that I ever used). The strange thing, that bald dude's avatar is png as well as the ones that are broken, I just loaded it manually, and it started to appear properly.. But trying it with others doesn't always work.
I recently just learned that It happens to break when konqueror redirects itself to http://sitename:80 even though none of the links have the port as part of the href. Once I remove the :80 from the url bar, and hit enter, the page reloads WITH working images.
Possibly related: bug 173950
assign
I get similar results all over the place. www.naturescapes.net, www.birdphotographers.net, www.fotocommunity.de, I can continue the list of broken rendered sites for quite some time - every now and then all images are shown, but often I have to hit reload to get all to show. My konqueror ist a SVN build as of today (2008-11-29).
*** Bug 177345 has been marked as a duplicate of this bug. ***
*** Bug 178312 has been marked as a duplicate of this bug. ***
David was nice again and provided patch which seems to work fine: http://www.davidfaure.fr/kde/kio_http_nodefaultport.diff
(In reply to comment #7) > David was nice again and provided patch which seems to work fine: > http://www.davidfaure.fr/kde/kio_http_nodefaultport.diff I'm wondering how this fixes it? The port I've had the problems with is port 80. Which is the default as far as I know. But konqueror keeps being told by kio to forward to sitename:80, even if sitename itself is port 80.
*** Bug 186358 has been marked as a duplicate of this bug. ***
*** Bug 188435 has been marked as a duplicate of this bug. ***
Bug 188435 has more details on this issue. In short some redirects in Konqueror are broken.
*** Bug 189110 has been marked as a duplicate of this bug. ***
This is fixed in trunk, not (yet) backported to 4_2_branch.
What a relief!
This doesn't seem to be fixed at least on www.birdphotographers.net. I haven't applied the default-port patch to 4.2 because it looked slightly risky. OTOH, I didn't exactly apply David's patch but a heavily modified version. It should do the same thing though. If somebody can confirm that things work with David's patch but not with the patch in trunk I'll have another look. Otherwise I'll assume that something else is broken because that seems more likely.
I think bug #153602 is a duplicate of this bug.
(In reply to comment #16) > I think bug #153602 is a duplicate of this bug. I think not, cause if you go from any microsoft.com page to another one and just hit "back" in konqueror the page will be shown correctly.
SVN commit 993441 by ahartmetz: - Proper handling of the situation that the server might not supply a way to determine the end of response body on a keep-alive connection - we need to prevent keep-alive then. - Some more simplification of the "do we need to disconnect from the current server for that new URL" logic. This includes removing the fugly m_isFirstRequest hack. - Completely rewritten cache, with short methods, some separation of concerns and whatnot. Also binary storage of metadata that will be updated often for better performance and smaller size. Only one file per resource instead of two. - Create cache files in the HTTP ioslaves, however serialize stat *updates* by doing them through the (now always running) cache cleaner process. - New cache file naming scheme. I asked Chris Mason (of btrfs) on IRC about generic Linux filesystems and directories with a few thousand entries. All mainstream filesystems can handle that, so we'll just hash the URL and not use subdirectories. - Add --file-info option to the cache cleaner. Very helpful for debugging. - Follow the RFC more closely wrt caching, among other things by using the Age header as a last resort to determine the original serve date of a resource. - Regression-ish: no compression - I'd rather not compress HTML and CSS than try to compress everything including images. Compression also seems to have interacted badly with updating file stats on disk. It will be possible to re-add compression, there is a reserved byte in the cache file header for that. If at all, only for text types please. - Move cache related header parsing into a helper method to shrink the huge readResponseHeader() method. We also skip processing that stuff if we are not going to need it. - Generally more data members were removed than added in http.h. Yay! BUG: 171256, 193181 M +935 -904 http.cpp M +62 -66 http.h M +426 -200 http_cache_cleaner.cpp M +8 -19 parsinghelpers.cpp M +1 -1 parsinghelpers.h WebSVN link: http://websvn.kde.org/?view=rev&revision=993441