The 'network' tab in dev tools records the header for the css GET as using req header with NOT spoofed user-agent. Some of the other files' (these all seem to be blocked js-files) req headers are with non-spoofed UA too, but they have that "Provisional headers are shown" with them; the css GET does not have that. Other files do have spoofed UA, as expected.
I assume CF uses UA to compute a part of those "clearance" cookies. This is because the CF "security-check" screen pops back up to block access when changing UA spoofing ON/OFF. If the non-spoofed UA is really sent out, the discrepancy might trigger the problem. Just a guess, though.
ps.
It would actually seem that all css-GETs are done with no UA spoofing. Only CF "cares" that those match; thus the initial problem.
v1.0.8.5 has officially been released. Could you please retest (once you verify that ScriptSafe v1.0.8.5 is installed) and let me know if the issue is still present? Thanks!
Hello.
(Chrome/51.0.2704.103) UA spoofing seems to cause some problems with CloudFlare. Problems occur when trying to load css styles for a CF-protected page "https://www.random.org" from "https://www.random.org/css/style.css". When accessing "https://www.random.org/css/style.css" manually, there are no problems and the css displays as a page of text, as expected. But when "https://www.random.org" tries to internally GET "https://www.random.org/css/style.css", console logs 403 for the GET.
The 'network' tab in dev tools records the header for the css GET as using req header with NOT spoofed user-agent. Some of the other files' (these all seem to be blocked js-files) req headers are with non-spoofed UA too, but they have that "Provisional headers are shown" with them; the css GET does not have that. Other files do have spoofed UA, as expected.
I assume CF uses UA to compute a part of those "clearance" cookies. This is because the CF "security-check" screen pops back up to block access when changing UA spoofing ON/OFF. If the non-spoofed UA is really sent out, the discrepancy might trigger the problem. Just a guess, though.
ps. It would actually seem that all css-GETs are done with no UA spoofing. Only CF "cares" that those match; thus the initial problem.