* TITLE/SUMMARY Modified image used on page is not updated between page reloads * STEPS TO REPRODUCE 0) Launch Apprunner 1) Execute the img-reload text case (temporarily located externally at http:// www.prometheus-music.com/gecko/img-reload/img-reload-desc.html --- you'll probably need to copy down the files to your local hard drive to execute it.) * RESULT - What happened While the foo.jpg image is replaced with a new image file between page loads, the original image (still in memory) is displayed, instead of the altered image. - What was expected Images to be checked for changes between page reloads. Otherwise, web designers will go crazy. ;) * REGRESSION - Occurs On Mac OS Apprunner (5.13.99 AM optimized build) Win32 Apprunner (5.11.99 AM optimized build [NT 4, Service Pack 3]) Linux Apprunner (5.11.99 AM optimized build) - Doesn't Occur On Mac OS Communicator 4.6 (99124 pre-RTM build) * CONFIGURATIONS TESTED - [Mac] Power Mac 8500/120 (233 Mhz 604e), 64 MB RAM (VM on; 1 MB of VM used), 1024x768 (Thousands of Colors), Mac OS 8.6 - [Win32] Vectra VL (233 Mhz P2), 96 MB RAM, 800x600 (True Color), NT 4.0 SP3. - [Linux] Vectra VL (266 Mhz P2), 96 MB RAM.

Pushed past necko landing...

Changing all Networking Library/Browser bugs to Networking-Core component for Browser. Occasionally, Bugzilla will burp and cause Verified bugs to reopen when I do this in a bulk change. If this happens, I will fix. ;-)

Deferred till Cache works in Necko.

[Courtesy notice from M9 testing: issue still present in 8.16.99 AM builds on Mac OS 8.6 & Windows NT 4.0 SP3.]

Deferring till cache lands. CC'ng fur.

hmmmm. A fix for this bug shouldn't depend on cache. I'm reassigning to myself, since I know Gagan has more than his share of bugs and I can check the perimeter of this problem pretty quickly. -pn

note: bug# 6627 for reference.

[Courtesy note: Bug still present on 9.23.99 AM build; tested Win32 on NT 4.0 SP3]

Uh. what happened to shift reload? The old world had a reload and a shift-reload. The shift-reload always reloaded the page and its contents. The reload tested the page and if had not changed, nothing was done. The individual elements on the page were not tested. If you needed your images 'refreshed' periodically, you had the option of using metatags to kick off a reload. I no longer see the reload/refresh on the menu list and shift-reload does a regular reload. -pn

Until the cache lands, there is no difference between default loading, reload, shift-reload and history-reload, i.e. back/forward buttons. Still, this should not be responsible for this bug since all netlib loads will effectively be the same as shift-reload until the cache is present, i.e. all HTTP requests go all the way out to the origin server.

Right. but the image request gets a NET_NORMAL_RELOAD from netlib, so it uses the image finds in the imglib cache. Shouldn't the image request get some different notification that the netlib retrieved the document from the server not the cache? -pn

hmmm. The NET_NORMAL_RELOAD is coming from a default value plopped into NS_NewImageNetContext(). This is where the policy breaks down. -pn

AFAIK, NET_NORMAL_RELOAD is dead, a legacy of the old netlib. Soon, there will be an equivalent flag added to the load flags of a channel, but it does not exist yet. When it is finally added, the front-end should set the flag based on whether or not reload or shift-reload is hit.

ok. Sounds like I should reset the milestone on this guy. I was making the assumption that the new flags were operational and I should temporarily reinterpret them into the old netlib states until the new cache was ready. Keeping the default to NET_NORMAL is probably less annoying than changing it to never use the image cache. Thanks for the info. BTW, if I can help test the new cache stuff, let me know. -p

Sorry; the actual test case (on marvin) was updated, but the temporary file posted on prometheus-music.com didn't mention holding the shift key.

*** This bug has been marked as a duplicate of 21036 ***

See bug 21137 to learn about new netlib flags.

QA Assigning to self.

Bulk move of all Cache (to be deleted component) bugs to new Networking: Cache component.

Verifying as "duplicate". Will verify this bug upon 21036's implementation.