Limit data-uri size to 32KB, falling back to relative url()#1190
Limit data-uri size to 32KB, falling back to relative url()#1190evocateur wants to merge 7 commits intoless:masterfrom
Conversation
Apparently there is some disagreement on the tangent of 40 between browser and node. 42 is the answer.
Although IE8 does support data-uris, it only does so with a limit of 32KB. It's a silly limitation, but a source of potential bugs. When the limit is exceeded, the data-uri() function will simply return a normal url() value with a relative path to the asset. One may pass --no-ie-compat to lessc to avoid this safeguard.
lib/less/functions.js
Outdated
There was a problem hiding this comment.
use tree.Quoted instead of tree.Anonymous then the url( will be quoted
There was a problem hiding this comment.
Makes sense, I'll push that shortly.
There was a problem hiding this comment.
On the other hand, quoting isn't necessary inside url() statements. I'm not sure why the browser version inserts the quotes, but neither syntax is breaking anything except conformity of output between the browser and CLI versions.
There was a problem hiding this comment.
I realized ieCompat should be in the evalEnv, not parseEnv.
|
regarding Quote - if you keep it anonymous and the path contains '"() then this will break because they are not escaped. |
|
Ah, yes, fair enough. Update incoming. |
|
@agatronic Any chance we can merge this in soon? Let me know if you have any remaining issues. |
|
@evocateur done. Note: I had to change it to make the tests pass with both types of seperators. |
|
Great, thanks! |
2 participants
Version 8 of every web dev's favourite browser limits
data-uris to 32KB. Since we're now in the business of embeddingdata-uris, we should avoid exceeding this.A new option for
lesschas been added,--no-ie-compat.By default, the corresponding env property
ieCompatistrue. Passing--no-ie-compator settingenv.ieCompattofalsewill avoid returning the fallbackurl()and always embed thedata-uri.The fix to
tan()was just something random I saw while trying to make bothnpm testandmake browser-testpass cleanly. I have no idea whynodeandphantomjswould decide those products differ (only the last significant decimal was off by one in PhantomJS), buttan(42)fixed it.