summaryrefslogtreecommitdiffstats
path: root/PROBLEMS
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 16:37:15 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2024-04-07 16:37:15 +0000
commitae5d181b854d3ccb373b6bc01b4869e44ff4d87a (patch)
tree91f59efb48c56a84cc798e012fccb667b63d3fee /PROBLEMS
parentInitial commit. (diff)
downloadlynx-ae5d181b854d3ccb373b6bc01b4869e44ff4d87a.tar.xz
lynx-ae5d181b854d3ccb373b6bc01b4869e44ff4d87a.zip
Adding upstream version 2.9.0dev.12.upstream/2.9.0dev.12upstream
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to '')
-rw-r--r--PROBLEMS258
1 files changed, 258 insertions, 0 deletions
diff --git a/PROBLEMS b/PROBLEMS
new file mode 100644
index 0000000..621ccc1
--- /dev/null
+++ b/PROBLEMS
@@ -0,0 +1,258 @@
+Lynx PROBLEMS file.
+
+Ideally you would never have to read this, but inevitably problems
+do arise. As implementation and installation problems become known
+they will be outlined in this file.
+
+---------
+
+ The install script will attempt to preserve your current settings in the
+ lynx.cfg file. You should review these (at the end of lynx.cfg) after
+ install. For instance references to lynx.browser.org for a startfile can
+ be removed, since that site is no longer maintained.
+
+---------
+
+ Some customized installs of OpenSSL are not detected by the configure
+ script. For example, with Redhat9, we did this to work around:
+
+ #!/bin/sh
+ KRB=/usr/kerberos
+ export LIBS="-L$KRB/lib -lssl -lcrypto -lgssapi_krb5 -lkrb5 -lk5crypto -lcom_err"
+ export CPPFLAGS="-I$KRB/include -I/usr/include/openssl"
+ export CFLAGS="-O -g $CPPFLAGS"
+ ./configure --with-ssl $*
+
+---------
+
+ The configure script check for IPv6 (--enable-ipv6) assumes that your build
+ machine can listen for IPv6 connections. It cannot determine if you want
+ to use it only for outgoing connections. If the configure script is
+ otherwise successful, it will warn about a nonfunctioning getaddrinfo()
+ function, and does not define these symbols needed to successfully compile
+ IPv6 support. You can work around this by by editing lynx_cfg.h:
+
+ HAVE_GAI_STRERROR
+ HAVE_GETADDRINFO
+
+---------
+
+ Control-Z on Unix can cause aberrant behavior. If you encounter
+ problems, use -restrictions=suspend to disable it, and only '!'
+ for escapes to shell (on VMS control-Z is unconditional 'Q'uit,
+ with no attempt to suspend the Lynx process only temporarily).
+ Control-Z on Unix often works better with slang.
+
+---------
+
+ Screen resizing can be a bit funny. If you resize the screen,
+ documents that have been cached will be out of whack. Any further
+ documents will look fine. You can reload documents to the current
+ window size with CTRL-R
+
+---------
+
+ If the configure script dies when it discovers that gcc is broken,
+ despite the fact there is a perfectly good ANSI cc in the PATH,
+ try setting the CC environment variable to your working cc compiler
+ before running configure: e.g.
+
+ setenv CC cc (or set CC=cc; export CC , depending on the shell)
+ ./configure
+
+---------
+
+ If you run into a problem compiling GridText.c, try omitting -O from the
+ compiler flags; it is a very big file and may not compile successfully
+ if you include optimization.
+
+---------
+
+ On a Sun system, the message:
+ "Alert!: Unable to connect to remote host"
+ will be displayed every time an attempt is made to access a host
+ other than localhost if Lynx has been built without the resolv
+ library and needed it, or with it and shouldn't have been.
+ Unfortunately, there's no way to check in the Makefile whether
+ -lresolv should be included in the LIBS="" list. What's necessary
+ depends on how that Sun is configured. To get the build right for
+ your SUN 3 or 4 OS, if you didn't have RESOLVLIB defined in the
+ Makefile define it and build Lynx again, or vice versa. Also, if
+ you have upgraded to the bind-8.1 or later library, you should try
+ changing -lresolv to -lbind.
+
+---------
+
+ On Sun systems when accessing sites that use multiple IPs for one URL,
+ you may experience a core dump. It's a bug in the resolver library,
+ which can be fixed by downloading & installing Sun's jumbo patch:
+ search Sun support information, e.g., Solaris newsgroups,
+ or see the following Lynx Archive items:
+
+ http://www.flora.org/lynx-dev/lynx-dev/9606/0422.html
+ http://www.flora.org/lynx-dev/html/month0497/msg01018.html
+ http://www.flora.org/lynx-dev/html/month0497/msg01064.html
+
+ The Sun `shelltool' and `cmdtool' terminals are stupid by default.
+ In order to get bold text to appear differently than inverse video,
+ the user should put this line in ~/.Xdefaults:
+
+ Term*boldStyle: Offset_X
+
+ From the `shelltool' man page, it seems that an analogous line
+ in ~/.defaults
+
+ /Tty/Bold_style "Offset_X"
+
+ ought to work just as well, but you may not get the desired
+ behavior until you modify your .Xdefaults file (and run `xrdb
+ ~/.Xdefaults'). Note also that there are other supported values
+ for the boldStyle resource/Bold_style default, all of which begin
+ with "Offset_". (helpful hint from kevin@traffic.den.mmc.com)
+
+---------
+
+ Directory browsing has been implemented for VMS, but there are no
+ plans to port additional DIRED support, because Lynx must handle files
+ as streams, and this precludes "serious" Directory/File Management on
+ VMS. Use a jumps file link to CSwing (sources or executables are
+ available from ftp://narnia.memst.edu), or define CSWING_PATH in
+ userdefs.h or lynx.cfg to invoke CSwing via the DIRED_MENU command.
+
+---------
+
+ If one switches between K)eypad "Numbers act as arrows" versus "Links
+ are numbered" in the 'o'ptions menu when the current document is a
+ DIRED menu in which links have been tagged, the tagging can be trashed
+ (so don't switch at such times 8-).
+
+---------
+
+ When "Links are numbered" is on, if a line is split on an anchor (to
+ obey right margin restrictions), and there is no space in the bolded
+ string such that the entire "[#]string" must be moved down, the "[#]"
+ becomes bolded, instead on only "string".
+
+---------
+
+ The Mosaic v2.5 hostlist uses HTML similar to that of Lynx's bookmark
+ file, but with </UL></HTML> at the bottom, such that it is not fully
+ compatible with Lynx's file. If you try to use the Mosaic file as if
+ it were a Lynx bookmark file, the </UL></HTML> will not be taken into
+ account and new links will be added below rather than above those end
+ tags. Instead, add a link to the Mosaic file in your Lynx file, and
+ to the Lynx file in your Mosaic file, so that you can access both files
+ with both clients.
+
+---------
+
+ SOCKSification and the -socks switch have not yet been integrated with
+ the slang library support.
+
+---------
+
+ There is an apparently broken version of select() in libcurses.a
+ of HP/UX 10.10. It also breaks tn3270, ncftp, emacs, and xemacs.
+ Using:
+ LIBS="-lc -lcurses -ltermcap \
+ ^^^
+ (i.e, adding -lc *before* the -lcurses) in the snake3 and snake3-slang
+ targets of the top level Makefile yields a usable image, but with
+ inappropriate video attributes on the Lynx displays (reverse video and
+ underscores on everything). Using "-lc -lHcurses" instead fixes the
+ ^^^^^^^^^
+ video attributes but then the arrow keys are messed up. - Donald S.
+ Teiser (dsteis01@homer.louisville.edu)
+ NOTE: If HP fixes the problem or you come up with a better workaround,
+ notify the lynx-dev@nongnu.org list.
+ Updated NOTE (1996-09-02): A patch reportedly is available from HP to
+ fix the select() problem, so that "-lc" is no longer needed, but
+ the curses glitch is not yet fixed, and you should still include
+ "-lHcurses".
+ Updated NOTE (1997-02-03): The problems reportedly are fixed with
+ patches PHCO_8086 and PHCO_8947 from HP.
+ Updated NOTE (1997-12-15): PHCO_8086 & PHCO_8947 are very old and are
+ no longer available. The current patch to install if running
+ under HP-UX 10.20 is PHCO_11342.
+
+---------
+
+ Lynx juggles variable abilities of curses packages or emulations to
+ display bolding and underlining simultaneously. This may fail if
+ Lynx thinks that your terminal, in connection with the curses package,
+ supports a capability which the terminal hardware or emulation does not
+ in fact support. Setting the right TERM environment variable, tweaking
+ terminfo or termcap files, or compiling with a newer version of ncurses
+ or slang may solve problems with missing highlighting or strange
+ characters appearing on the screen. Also, for a mono terminal,
+ make sure "show color" is not set to ON in the Options Menu.
+
+ The Wyse 50 and older TeleVideo terminals, among others, are
+ "magic cookie" terminals. This means that display attributes like
+ reverse, blink, underline, etc. work in a bizarre way that makes them
+ difficult to program. You may see extra spaces scattered around your
+ screen (separating different sorts of highlight); or sections of the
+ screen may be unexpectedly highlighted.
+ There is a workaround which works by restricting the terminal to a
+ single standout attribute (e.g., normal and reverse, but no others).
+ Implementing the workaround is specific to your curses implementation.
+ Most versions of curses use one of two terminal databases, called
+ "termcap" and "terminfo". Updating these databases is system-specific.
+ New databases should be available from the vendor or other sources.
+ For the Wyse 50, try
+ <URL: http://www.wyse.com/text/custserv/faq/wy50faq.htm>;
+ extract the "wy50" (NOT "wy50-mc") entry and use that in place of the
+ existing one. See `terminfo', `infocmp', `tic' etc. man pages if
+ necessary.
+ Alternatively, compiling Lynx with the slang library may avoid problems
+ with these terminals.
+
+ The Sun console driver (aka wscons(7)) implements "reverse" and "bold"
+ as "reverse", causing confusion where Lynx uses the distinction between
+ the two to convey information. Lynx tries to detect this automatically,
+ but if it fails (for instance, you are running under "screen"), try
+ setting the -noreverse commandline option.
+
+---------
+
+ On VMS, Lynx, and other TCP-IP software, have been experiencing chronic
+ problems of incompatibilities between DECC and MultiNet headers whenever
+ new versions of either DECC or MultiNet are released. The Lynx build
+ procedure for VMS and a maze of spaghetti #ifdef-ing in tcp.h of the
+ libwww-FM had previously been successful in dealing with this problem
+ across all versions of MultiNet and of DECC, VAXC, and Pat Rankin's
+ VMS port of GNUC, but are now not 100% successful. If you get compiler
+ messages about "struct timeval timeout" having no linkage, add that
+ declaration immediately below the inclusion of ioctl.h for MultiNet in
+ tcp.h (by deleting the "#ifdef NOT_DEFINED" and "#endif /* NOT_DEFINED */"
+ lines):
+ [...]
+ #include "multinet_root:[multinet.include.sys]ioctl.h"
+ struct timeval {
+ long tv_sec; /* seconds since Jan. 1, 1970 */
+ long tv_usec; /* microseconds */
+ };
+ [...]
+ If you get compiler warnings about incompatible multinet_foo()
+ declarations, delete those where indicated in tcp.h. For the most
+ current versions of MultiNet, you can modify tcp.h to use the DECC
+ socket and related headers.
+
+ On VMS, the ftp function does not work with SOCKETSHR 0.9D and NETLIB
+ 2 (NETLIB 1 may work). This is because the functions getsockname()
+ and getpeername() within SOCKETSHR make incorrect calls to the NETLIB
+ functions. This results in zeroes being returned for part of the local
+ IP address. Since ftp sends this IP address to the remote end, the
+ remote server ends up sending a file back to a non-existent address.
+ Andy Harper (A.HARPER@kcl.ac.uk) has fixed these problems in the
+ SOCKETSHR 0.9D sources and offers the fixes as:
+ http://alder.cc.kcl.ac.uk/fileserv/zip/socketshr_src_09d-2.zip
+ ftp://ftp2.kcl.ac.uk/zip/socketshr_src_09d-2.zip
+
+---------
+
+ On VMS, to build an SSL-capable version, lynx and the ssl library
+ e.g., OpenSSL, must be built using the same network library. If you
+ build OpenSSL without specifying the network library (the 5th parameter
+ of the makevms.com script), it will guess, possibly not the one you
+ intended. We have tested only the UCX configuration -TD (2002/9/15).