prototype for idle() in <unistd.h> (which i thought was against namespace
and sensibility guidelines, but...)
* consistently use xsignal() instead of signal(). we get known behaviour
in all cases (SA_RESTART), which is good for some borken foreign systems.
* remove signal.h from most files; it's unnecessary now
* fetch_url(): use `long chunksize' instead of ssize_t; it's more portable, and
we're setting chunksize with strtol() anyway
* xsignal(): only use SA_RESTART if it exists. SunOS 4.x doesn't have it
but has the inverse (SA_INTERRUPT). the original function i was inspired
from had this support (lib/signal.c, W. Richard Stevens' `UNP 2nd ed Vol 1').
* remove <termios.h> from util.c; it should be unnecessary now
prototype for idle() in <unistd.h> (which i thought was against namespace
and sensibility guidelines, but...)
* consistently use xsignal() instead of signal(). we get known behaviour
in all cases (SA_RESTART), which is good for some borken foreign systems.
* remove signal.h from most files; it's unnecessary now
* fetch_url(): use `long chunksize' instead of ssize_t; it's more portable, and
we're setting chunksize with strtol() anyway
* xsignal(): only use SA_RESTART if it exists. SunOS 4.x doesn't have it
but has the inverse (SA_INTERRUPT). the original function i was inspired
from had this support (lib/signal.c, W. Richard Stevens' `UNP 2nd ed Vol 1').
* remove <termios.h> from util.c; it should be unnecessary now
it's more portable and more obvious
* remove the mkgmtime() && HAVE_TIMEGM stuff:
a) why should netbsd have to define HAVE_TIMEGM to compile cleanly?
b) foreign compiles of ftp should just be linked with working
timegm function
a more portable version of this ftp client will be released as a 3rdparty
product; no use polluting our code with half-baked attempts...
* abort_remote(): replace borken MIN(4,BUFSIZ) with just BUFSIZ; it
should have been MAX(4,BUFSIZ), but it's probably safe to assume that
BUFSIZ is at least 3... (fix from simonb)
* auth_url(): use the correct variable when calculating a buffer size.
add lots of comments about how to size up the buffers, and add extra
checks to hopefully ensure that there won't be an overflow (unless
someone modifies the length of the sprintf()s).
* as part of the above, slightly rework the way the `*' bar is calculated.
also fixes a display bug when > 160 stars were needed to be printed.
the maximum progress bar width at this time is 256.
* remove some code that checks the port that was #if 0-ed out as part of the
ipv6 migration; it's not going to be used again.
* document the above three commands
* rototill the way the sndbuf and rcvbuf work. remove resetsockbufsize()
* use the appropriate socket buffer size as the size of the buffer that
the read()/write() loops use. speeds up things in some cases.
list of target nodes returned by Make_ExpandUse(). We have to search the
cohorts explicitly while iterating through the tree. So, tweak CompatMake()
to do this.
can (i.e., everything except environment variables, which aren't
stored in hash tables).
While we're here, inline the body of VarDelete into Var_Delete since
it's the only caller and it's just simpler that way when v->name can
share storage with the hash entry and may not need to be freed
separately.
Speeds up the infamous libc build benchhmark maybe 1% on PIII, 4% on
alpha pc164
Suggested by Perry Metzger.
Build a list of `cohorts' as before, but do *not* link each one into all the
parent nodes; instead, copy the `cohort' lists into the stream of targets to
be built inside Make_ExpandUse(). Also do the attribute propagation as a
separate pass after parsing.
This eliminates several O(n^2) algorithms.
passing in the LstNode of the child being inspected. Shaves off another few %,
particularly when there are long child lists containing $ expansions (e.g. in
libc).
* Treat search patterns as BREs only, per 1003.2 and XCU5.
* Do not perform any special treatment when an unqoted * or @ is given as
the first character of the search pattern.
Addresses PR bin/8269 by Chris Demetriou.
connection destination, hoping this to help ftpd's behavior with
scoped IPv6 addresses.
I'm not sure if it is the right way, but it is the best way available to us.
LPRT or EPRT command gives no information about which interface (or scope)
to be used for new data connection.
ftp(1): On data connection establishment, warn if scoped address is used.
If peer (ftp daemon) does not handle scoped address, data connection
may not work right.
This seems to be sort of protocol spec hole, not implementation issue.
a malloc'ed pointer and it tries to realloc(3) it if it had to grow it
before. su(1) gave it a pointer from the stack which caused realloc to
core dump.
- just display the hostname:port of the proxy url, rather than the full url.
this prevents someone `shoulder surfing' a proxy username/password
in $http_proxy. [suggested by perry]
- compact verbose notes for http fetchs; now displays
(via host:port, with authorization, with proxy authorization)
with each component being optional.
(and a couple introduced with the ipv6 mods...)
- don't override host with the canonical name; this prevented fetches from
http/1.1 virtual hosts from working if the virtual host was a CNAME.
[noted by bernd]
- call freeaddrinfo() if res was built with getaddrinfo()
some of them are caused by the kvm_getfiles() implementation being
TOTALLY BROKEN for dead kernels. Print out the KVM descriptor's
error so that error causes are more easily diagnosed.
