1364722 - Unnecessary requests re-sent for a cached image with only the hash at the end of the URL changed

Status: RESOLVED DUPLICATE of bug 1027106

(Core :: Graphics: ImageLib, defect)

Description

[Mardeg]

Attachment: SVG file for testcase

Tested in latest Nightly on Linux 32-bit, in the devtools network monitor I'm seeing the same URL request being re-sent when javascript is used only to change the part of an SVG image address after the # (it's a sprite sheet of a dozen images with <view>s of each.

STR: 
1. Open the attached testcase and devtools network monitor.
2. Choose a version of the image by host (please don't hammer imgh.us)
3. Click the image to see the next sprite
4. Check the network requests.

Actual behaviour: Each time the image is clicked the same URL is re-requested over the internet.

Expected behaviour: There should be a check done when the img .src attribute is set and no part before the # has changed, to prevent an unneeded round trip to a server. Chromium does this check and the image changes instantly in response to clicks.

Comment 1

[Mardeg]

Attachment: img src testcase

Note the #dc01 at the end of the img src URL ascends to #dc02 etc. for the next sprite up to #dc12 and cycles back to the beginning.

Comment 2

[Nathan Froyd [:froydnj]]

I am 95% positive that we have tests that depend on the fragment behavior not hitting in the image cache.

Comment 3

[(no longer active)]

Daniel, do I remember correctly that this is a known issue with SVG images?  I remember running into a similar issue when working on tab audio indicators where we also use SVG images with URLs with hashes at the end where we refetch the image similarly...  I can't find the bug# right now.

Comment 4

[Mardeg]

Preserving a comment here from bug 1347376 comment 9

> it occurs to me the best way to make LoadImage() faster is not to call it at all.

Comment 6

[Daniel Holbert [:dholbert]]

tnikkel found the correct dupe-target -- thanks!

(Per bug 1121693 comment 8, this isn't SVG-specific -- it happens for whatever.jpg#dc01, whatever.jpg#dc02, etc. too -- though it's more likely that someone would want to use a #ref with a SVG image.)

Comment 7

[Mardeg]

Attachment: SVG file for testcase #2

Comment 8

[Mardeg]

Attachment: Caption file for testcase #2

Comment 9

[Mardeg]

Attachment: OGA file for testcase #2

Comment 10

[Mardeg]

Attachment: MP3 file for testcase #2

Comment 11

[Mardeg]

Attachment: testcase #2

Shows how crucial having no delay from a "load" cycle is.