loop even when the timeout is zero.

when the extension is natively supported by Cryptlib.  This is to work
around the apparent impossiblity of adding multiple DNSName entries to the
Subject Alt Name field.

Sorry DigitalMan, you'll have to compile the Win32 libraries again.  :-)

The Win32 lib and header still needs to be updated.

The Win32 bits haven't been tested yet, and I plan to update to 3.4.4 in a
few minutes.

This was needed for libmozjs 1.8.5 to compile with GCC (g++) v6.2.1. Hopefully
this works with clang (FreeBSD) too.

TERM is set to "syncterm", this is a global Program Setting, so the best
thing to do is use the SyncTERM termcap or terminfo entry.

Otherwise, you'll need to select a working TERM value from what's available
on the remote system after logging in.

This results in *.vcxproj files not being detected as ascii since they have a
UTF-8 BOM at the beginning (and are therefore not ASCII).

Use perl to strip CRs from crypt32.vcxproj to allow the patch to apply cleanly.

be shared across OSs.

Make the integrated mingw32 Cryptlib build work on my system.

use an older Cryptlib anymore.

The access paramter to the ACL is bogus, but it works.

on an ARM system.

Also, add a work-around the the GCC visiblity pragma usage to permit building
with Clang 3.3

since we're including it in a .so.

the configure script supports aside from armv7).

Also, paw through the gcc -v output to discover if it was configured to use
the VPD and hard float or not.  This makes Spidermonkey build on Raspberry Pi
with the hard float Raspian image.

child to avoid bad interactions with the SDL atexit() handlers.

3.4.2 will not need this patch.

down GC issues.

generated config files rather than patching the sources.

(As a side note, there is an Mk/bsd.gecko.mk!)

pthread headers.  Fixes problem linking JS on FreeBSD 9.0-RC1.

Don't pass -o and -u to unzip... just -o (-u contradicts -o)

This is because the Cryptlib makefile doesn't use $(MAKE) for submakes and
instead hardcodes the string "make" since gmake -C sets MAKEFLAGS to "w",
this ends up breaking builds on systems where the first make in the path
doesn't support the non-standard -w option.