(Resending in plain text due to the HTML scrub mess I caused. SORRY!) ----------------
Hello, fellow Fossil users! This is my first message on this mailing list, and I'd like to share my log of today's struggle. As the subject says, I just found some weird action of Fossil. To make long story short, I tried to run Fossil SCGI server with aid of nginx. So I put few of my Fossil repositories under `/usr/local/fossil_repositories`, gave ownership of `www:www` to that directory, and wrote following configuration files: (I referred Fossil wiki article [1] while doing this.) ## /etc/rc.conf # run nginx upon booting nginx_enable="YES" # run Fossil as SCGI server daemon upon booting. fossil_enable="YES" fossil_user="www" fossil_port="53080" fossil_directory="/usr/local/fossil_repositories/" fossil_proto="scgi" fossil_listenall="" # Leave this empty!! fossil_notfound="http://fossil.my.example.com/index.html"; ## /usr/local/etc/nginx/nginx.conf: Default nginx config file on FreeBSD http { ... server { listen 80 default_server; server_name fossil.my.example.com; location / { root /usr/local/www/public_html; index index.html; # Display the list of ALL Fossil repositories. } location /my-homepage/ { include scgi_params; scgi_pass localhost:53080; scgi_param SCRIPT_NAME "/my-homepage"; } ... } } According to service init script, the command got invoked by rc script is: `/usr/local/bin/fossil server -P 53080 --scgi --localhost --notfound http://fossil.my.example.com/index.html /usr/local/fossil_repositories/`. After configuring, when I headed http://fossil.my.exmple.com/my-homepage/ I got redirected to http://fossil.my.example.com/index.html, which is clearly `fossil_notfound` fallback. So I started digging into source code. What I found while scheming through main.c is that I can give `--httptrace` command line option to any of subcommands. So I added that option, ran Fossil command above manually, and got following output on another web browser request: ... # cgi: REQUEST_URI = [/my-homepage/] ... # cgi: SCRIPT_NAME = [/my-homepage] ... # cgi: PATH_INFO = [/] ... Seems nothing wrong, since the relevant source code [2] says what Fossil requires is `SCRIPT_NAME`. Still in question, I ran Fossil HTTP server instead of SCGI, and made direct access like this: http://fossil.my.example.com:53080/my-homepage/. This time, I got WORKING Fossil web UI and following trace output: ====== SERVER pid 93606 ======= ... # cgi: REQUEST_URI = [/my-homepage/] # cgi: SCRIPT_NAME = [] # cgi: PATH_INFO = [/my-homepage/] ... # repository: [/my-homepage.fossil] # new PATH_INFO = [/] # new SCRIPT_NAME = [/my-homepage] ====== SERVER pid 93607 ======= ... # cgi: REQUEST_URI = [/my-homepage/index] # cgi: SCRIPT_NAME = [] # cgi: PATH_INFO = [/my-homepage/index] ... # repository: [/my-homepage.fossil] # new PATH_INFO = [/index] # new SCRIPT_NAME = [/my-homepage] # login: [nobody] with capabilities [gjor] Then I got an idea; how about omitting `SCRIPT_NAME` completely? So I amended nginx.conf like this: location /my-homepage/ { include scgi_params; scgi_pass localhost:53080; scgi_param SCRIPT_NAME ""; # EMPTY! } Finally, upon heading http://fossil.my.example.com/my-homepage/, I got WORKING Fossil web UI again. Problem solved. At least I found a detour, I still don't know why this happened though I followed Fossil wiki's instruction. Can someone identify the reason please? p.s. Actually, the true reason of this message is to answer Jon's question [3]: > How do you run `fossil server` in verbose/debug mode and/or have it log out > detailed info? Use `--httptrace`! [1]: http://www.fossil-scm.org/fossil/doc/tip/www/server.wiki#scgi [2]: http://www.fossil-scm.org/fossil/info/76a4af25a1df3fa6c112d09e291c4bba140df662?ln=864 [3]: http://lists.fossil-scm.org:8080/pipermail/fossil-users/2014-March/015478.html -- Ardie Hyun Hwang ---------------- email: i...@ardiefox.me mobile: +82-10-I-AM-ARDIE _______________________________________________ fossil-users mailing list fossil-users@lists.fossil-scm.org http://lists.fossil-scm.org:8080/cgi-bin/mailman/listinfo/fossil-users
