The proxy patch applied in 1.11.19-3 causes weird breakage here; this is a FC3 system with cvs locally rebuilt and updated from FC devel CVS. I have a local squid running, and http_proxy set to http://localhost:3128 in the environment. Now, all my cvs operations that use anonymous pserver fail with messages like: cvs [update aborted]: proxy server localhost:3128 does not support http tunnelling Huh? Why is it trying to use a HTTP proxy in the first place? I requested a CVS pserver operation, not HTTP traffic. The error message is confusing as squid most certainly supports HTTP proxying, and it works fine here using wget, curl, $younameit. Ethereal dump attached, and squid logs say: ==> /var/log/squid/access.log <== 1110118551.721 0 127.0.0.1 TCP_DENIED/403 1336 CONNECT cvs.cvshome.org:2401 - NONE/- text/html ==> /var/log/squid/store.log <== 1110118551.721 RELEASE -1 FFFFFFFF 652777049644E74A1CCE533ABF123F48 403 1110118551 0 1110118551 text/html 1044/1276 CONNECT cvs.cvshome.org:2401 I suggest turning this feature off by default, as it seems to be implemented in a way that causes backwards incompatibility and breakage. The changelog refers to #144297 which I'm unable to access.
Created attachment 111713 [details] Ethereal dump of failed "cvs update"
Could you join please list of your env variables? Do you have set "HTTP_PROXY", right?
I removed check of HTTP_PROXY. Now it works only with CVS_PROXY.
That sounds like a good idea to me, thanks. 1.11.19-4 looks ok. FWIW, here are my related environment variables: $ env | grep -i proxy http_proxy=http://localhost:3128 ftp_proxy=http://localhost:3128 no_proxy=localhost,bobcat.mine.nu,127.0.0.1