| Age | Commit message (Collapse) | Author |
|---|
|
authentication:
- Combine the http and https usage formats into a single http[s] format
to make it more concise.
- In the AUTO-FETCHING FILES section of the ftp(1) man page, mention
that specifying "user" and "password" with HTTP and HTTPS URLs will
log in using Basic authentication (if http_proxy is not defined).
- When compiled with -DSMALL, fix ftp(1) usage so that
"[user:password@]" is not shown for http[s] since Basic authentication
is not supported with -DSMALL.
Done with a lot of discussion with and help from jmc@ (thank you!).
ok deraadt haesbaert jmc
|
|
spotted by bluhm.
ok bluhm.
|
|
RFC 3986. This allows the following idiom in ftp:
ftp http[s]://user:pass@host/file
With some pointers from halex a lot of testing and feedback from lteo,
thanks a lot.
ok lteo@
|
|
looks on the man page.
ok haesbaert jmc henning
|
|
of the connection. This is useful for testing ftp(1) over VPN tunnels.
This -s flag is present in the other BSDs, including OS X.
All work was done by Lawrence Teo, thanks (-:.
ok myself mikeb
|
|
|
|
advice and ok martynas@
|
|
|
|
|
|
responses. Yes, they are violating the rfc's. Yes they do exist anyway.
Also fix a memory leak when url_encode fails to malloc, by simply
err'ing out if so.
feedback and ok phessler@
|
|
who asked me to commit because he is just running out of the door.
|
|
ok phessler@
|
|
|
|
- if a remote file by the name '-' is retrieved, that does not imply
it should go to standard output...
- make -o '' reset any previous -o action
- properly handle multiple -o 's
ok phessler@
|
|
found by Alexander Schrijver
OK halex@
|
|
a download would stall at the end of the file for no apparent reason.
ok sthen@, phessler@
|
|
doesn't support resume, we restart the download from the beginning, like
all other browsers.
Diagnosed by sthen and halex, comment from sthen
OK sthen@, halex@
|
|
gcc4 gets upset about.
based on suggestions from miod@, ok millert@
|
|
for auto-completion of filenames containing whitespace was added.
This bug was tracked a decade ago in NetBSD as pr 7014.
ok deraadt
|
|
|
|
Before this change whitespace wasn't escaped during completion which
caused the ftp command to see multiple arguments where a single path
was expected. Based on similar commit made to NetBSD's ftp in 1999.
help and ok uwe tedu deraadt
|
|
ok jsg millert
|
|
transfers. makes output between all transfers consistent; handles
SIGINFO for non-ftp transfers too; shows statistics at the end:
7303400 bytes received in 96.00 seconds (74.29 KB/s)
tweak / ok halex@, sthen@.
|
|
from the URL. In cases where the URL supplied on the command-line returned
an HTTP redirection, ftp(1) was changing the filename to one derived from
the redirected URL. Change this to always use a name derived from the
command-line URL.
Avoids unexpected behaviour with URI-encoded redirection URLs as seen
by matecocido on misc@, and avoids surprises when the redirection is
to an unexpected filename. No change when -o is used.
ok martynas@ deraadt@ "I totally approve" halex@
|
|
|
|
based on a conversation with martynas@; diff tweaked by jmc@ and martynas@
ok jmc@, martynas@
|
|
- urls passed to ftp containing special characters or unsafe
characters (like, spaces, <>"#{}|\^~[]`%) work
- redirects containing special characters or unsafe characters
(like, spaces, <>"#{}|\^~[]`%) work
reported & tested by ian@ (can fetch a distfile from sourceforge now)
feedback + ok tedu@ & sthen@
|
|
fetchable with -o. so point to it too. requested by chris@
ok sthen@. better halex@
|
|
fetchable, if -o outfile is passed. outfile will be used as a local
filename
- fix a bug where 'no file after host' code path never got entered;
consider no file after dir invalid; as code intended
proxy help&ok halex@; testing(including proxies,pkg_add)&ok sthen@;
looks good to millert@
|
|
No binary change.
|
|
needed not to make a status of 200 (OK) fall through to the redirect
handling.
This was not noticed since the later code would ignore the isredirect
flag unless it also received a "Location: " header.
ok martynas@
|
|
file transfers in all cases
- do it a bit earlier and use HTTP/1.0 if we won't send the range header
- change resume -> restart_point where it is intended
ok halex@, millert@
|
|
interrupt will do the standard abort procedure per rfc959. second
interrupt will force the abort, and close the connection.
done after email from halex@ couple of months ago. liked by many
"I like this a lot" marco@. ok halex@, millert@, phessler@
|
|
|
|
"looks sensible" deraadt@, "fine by me" martynas@
also removed an indeed misplaced/outdated comment per martynas@ request
|
|
actually know that that login is not provided in url; and we should
guess it. fixes a bug reported by halex@ where it tried to login
as 'ftp' after unsuccessful logins; ok theo, halex@, krw@
|
|
- don't spam ftp server with four anonftp logins; if ftp_login has
already failed in setpeer we give up. also makes install script
nicer; req'd by theo
- fix autologin = no case which never really worked in fetch since
it did not check for autologin before logging in...
part 1: ok theo, krw@
part 2: "looks right" millert@, ok sthen@
|
|
|
|
small stuff. make it a fetcher. shrinks quite a bit
agreed by millert@, krw@
ok theo, sthen@
|
|
install media if, e.g. ipv6 is not present
discussed with theo, sthen@, millert@, todd@
|
|
some space
discussed with theo, todd@, millert@
tested and looked over by sthen@
|
|
|
|
|
|
-T and just show titles by default for non-verbose transfers;
discussed with, ok theo, sthen@
|
|
the text segment slightly (though due to padding it does not change the size
of the binary). ok martynas@
|
|
server closes connection during any get/put transfer. it dereferences
cout, which will be NULL, oops
originally reported on misc@. pirofti@ can't reproduce it again
since the server has been fixed
tested by simulating the very same failure
sthen@ and pirofti@ agree that my analysis is correct
ok sthen@
|
|
show which filename is being used. ok krw, mdoc repair from jmc
|
|
so that recursive transfers don't cause a loop. problem reported
and tested by Jesus Sanchez. string compare inlining and looks
good to millert@
|
|
dante instead. "makes sense" sthen@, ajacoutot@, "no objections"
millert@; "i agree" jakob@. tested by Simon Kuhnle and myself
|
|
with strtonum
|
