We have more reports about this issue, i will dupe them to this bug because it contains the best description. Reporter: Are you willing to help us with creating a Cache logfile ? You would have to use a Aurora build that has logging of the cache enabled and creating the log. This logfile can get very big depending on how long it will take to reproduce the issue after clearing the disk cache. bug 808532/bug 771832 is related
I request to block for Firefox 18. I know that it's to late for blocking and I'm not sure if this is already fixed in FF18 but this looks like a really big issue to me. Reporter or someone from the dupes: Could be please run http://www.mozilla.org/de/firefox/beta/ and test it ? Please clear the disk cache once to be sure that your cache is in a good state.
I tested beta version for an hour and it seems to work.
ooops, it did it again, when I started to work at home.
Happy new Year and thank you for the positive response. Sorry about writing back so late. Due to new years. I actually have to work tomorrow again, but I'll do some intensive testing on the laptop tomorrow evening with the Aurora build and will report back. Disabling the AntiVirus didn't do the trick either. In the mean time some people reported on the support thread that cleaning the cache completely with an external tool got rid of the problem entirely. From the surf behavior of my wife revisiting the same pages over and over again, my guess is that either the initial safe of cache data fails or when the cache starts to purge old data it purges too much (like a whole sub branch like the "last" subdirectory in the path, which also contains data that is not to be purged). This however is just an educated guess and might of course be wrong. I hope the Aurora Build Data is going to enlighten the problem further. I have a few questions about this though: * Can you supply a direct link to the build I am supposed to use, please? * Will log in data (password etc.) be part of the log data (some of the errors actually happen on gmail or google+, so these are one of the best test sites for this)? Thank you for your replies.
Either one of these should give you the latest beta (Firefox 18). http://www.mozilla.org/en-US/firefox/all-beta.html http://ftp.mozilla.org/pub/mozilla.org/firefox/releases/latest-beta/ The latest Aurora build (Firefox 19) can be found here: http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-mozilla-aurora/ See also: http://www.mozilla.org/en-US/firefox/channel/ You can also delete the cache folder manually. You can find the location of the cache on the about:cache page.
I was just able to check one more case of a page loading incomplete in my work break with the stable release (I won't get to test any other until tonight). In this case pressing CTRL-F5 brought the page back fully. Before I checked like last time if the missing file would be found in about:cache but not physicallly. Trying to access the the html code of the page to determin the missing .css file, I already got an error for firefox loading the html code. Curiously the error must've occured DURING the page load, because the html part was clearly visible. Tracing back the cache entry I found that it was saved in cache directory .../2/12. Browsing to this directory I found that *all* subdirectories containing cached data were missing. There were only the 4 cache index files in the directory. Upon pressing CTRL-F5 the cache subdirectories reappeared. If this happens every time, It would explain why the effect is so drastic off an on. At the time I checked for the SWF there were other cache directories present, but maybe the cache just filled up again to a degree and it also had the previous purge happening as I could observe now. The people at the support forum wanted me to emphasize the fact, that after uninstalling the Samsung support center software, the problems with Firefox seem to be permanently gone for some users. I don't know if this is relevant to this bug though. Involvement of 3rd Party Software is possible, but who says this problem is limited to only one single program? I will pursue this possibility if testing with debug software won't yield any usuable results.
Oh and before I forget it, Google Chrome seems to have the same kind of issues as firefox.
How do I enable logging of caching and obtain the logs? I tried to look for a reference but didn't find any so far. Any hint would help.
> Reporter: Are you willing to help us with creating a Cache logfile ? > You would have to use a Aurora build that has logging of the cache enabled > and creating the log. This logfile can get very big depending on how long it > will take to reproduce the issue after clearing the disk cache. I have downloaded an Aurora build, however I don't know how to activate the logging capability there. Is that build already enabled for creating cache logs or would a custom build be required? Sadly I don't have any system setup to compile firefox right now, so that is out of the question so far. I am willing to test. The Aurora build also showed the same errors.
we want to help to identify the problem
it is annoying that you have two volunteers wishing to help and there is no response from developers :-(
I was able to build Aurora from source and checked various options. I believe though after looking at the source that the aurora nightly build should be sufficient for log generation. I was able to turn the logging on with the downloaded build and able to reproduce the problem. Sadly I was doing something more sensitive on e-bay, so I am reluctant to share this log file without looking at it first myself. In the mean time the Thread on the board seems to confirm, that the workaround by uninstalling one part of the Samsung Software seems to do the trick getting rid of the error. So this might be a software combination problem after all. To turn on logs, you need to create a batch file (.bat) which fill with (see http://www.mozilla.org/projects/nspr/reference/html/prlog.html for reference) : set NSPR_LOG_MODULES=cache:4 set NSPR_LOG_FILE=mylogfile.txt <absolute-path-to-firefox>\firefox.exe Replace the Path and maybe add a Path to the NSPR_LOG_FILE setting and then run the batch file. Erase cache and surf until the occurs. Best is to remember the URL it happened with. Close Aurora and you have the log. One strange phenomenon that I had was, that after closing Aurora the log file had NULL Bytes in the first X kb of the file, which prevented me of viewing the log in a normal text editor at first. After cutting the NULL Bytes away in a hex editor, all was good.
>In the mean time the Thread on the board seems to confirm, that the workaround by >uninstalling one part of the Samsung Software seems to do the trick getting rid of >the error. Which part ? Sorry that I somehow missed your comments but note that I'm not a developer.
I was referring to this Thread: https://support.mozilla.org/en-US/questions/941815 Current workaround is: 1) Removed Samsung SWUpdater and Support Center 2) Used CCleaner to fully clear all internet cache's 3) Rebooted 4) Optionally reinstall SW Updater Seems to work for at least 3 people now.
Bug 808532 has been present in 18,19, now fixed in all affected versions. Bug 771832 has been present in 4.0 up to 19, now fixed in versions 17 to 19. I think that while you were using a browser version affected with one of these bugs, a cache entry w/o the data file has been created. Condition to get rid of affect of those two bugs is to clean the cache. We had to land bug 812483 just for those affected versions to avoid the load error w/o forcing a silent or user initiated cache cleanup to fix the problem... To strength the theory, was the profile ever running a beta/aurora/nightly build of Firefox? (In reply to Leos Literak from comment #7) > ooops, it did it again, when I started to work at home. With what version exactly (about:buildconfig)?