Call stacks reported by valgrind:
==14702== at 0x80DD791: read_from_socket (socket.c:945)
==14702== by 0x8104D0C: read_more_http_data (http.c:1180)
==14702== by 0x81052FE: read_http_data (http.c:1388)
==14702== by 0x80DD69B: read_select (socket.c:910)
==14702== by 0x80D27AA: select_loop (select.c:307)
==14702== by 0x80D1ADE: main (main.c:358)
==14702== Address 0x4F4E598 is 56 bytes inside a block of size 81 free'd
==14702== at 0x402210F: free (vg_replace_malloc.c:233)
==14702== by 0x812BED8: debug_mem_free (memdebug.c:484)
==14702== by 0x80D7C82: done_connection (connection.c:479)
==14702== by 0x80D8A44: abort_connection (connection.c:769)
==14702== by 0x80D99CE: cancel_download (connection.c:1053)
==14702== by 0x8110EB6: abort_download (download.c:143)
==14702== by 0x81115BC: download_data_store (download.c:337)
==14702== by 0x8111AFB: download_data (download.c:446)
==14702== by 0x80D7B33: notify_connection_callbacks (connection.c:458)
==14702== by 0x80D781E: set_connection_state (connection.c:388)
==14702== by 0x80D7132: set_connection_socket_state (connection.c:234)
==14702== by 0x80DD78D: read_from_socket (socket.c:943)
read_from_socket() attempted to read socket->fd in order to set
handlers on it, but the socket had already been freed. Incidentally,
socket->fd was -1, which would have resulted in an assertion failure
if valgrind hadn't caught the bug first.
To fix this, add a list of weak references to sockets.
read_from_socket() registers a weak reference on entry and unregisters
it before exit. done_socket() breaks any weak references to the
specified socket. read_from_socket() then checks whether the weak
reference was broken, and doesn't access the socket any more if so.
This reverts src/{network,sched}/connection.c CVS revision 1.43,
which was made on 2003-07-03 and converted to Git commit
cae65f7941628109b51ffb2e2d05882fbbdc73ef in elinks-history.
It is pointless to check whether (c == d && c->id == d->id).
If c == d, then surely c->id == d->id, and I wouldn't be surprised
to see a compiler optimize that out.
Whereas, by taking the id as a parameter, connection_disappeared()
can check whether the pointer now points to a new struct connection
with a different id.
test/align.html: from www.czech-tv.cz
test/css/idnes_mail.html: apparently from idnes.cz
test/erreurs_en.htm: unclear origin
test/javascript_broken.html: presumably from www.hotjobs.com
test/poocs.net.html: presumably from poocs.net
I did not find in the source tree a licence to distribute any of these.
Samba 3.2.0 switched to version 3 of the GNU General Public License,
which is not compatible with GPLv2 used in ELinks. Perhaps we can
eventually relicense ELinks to "GPL version 2 or later", or resolve
this in some other way. Until then, warn distributors of binaries
about the conflict.
ELinks attempted to display a message box on file_download.term, but
it had already closed that terminal and freed the struct terminal. To
fix this, reset file_download.term pointers to NULL when the terminal
is about to be destroyed. Also, assert in download_data_store() that
file_download.term is either NULL or in the global "terminals" list.
Reported by أحمد المحمودي.
(cherry picked from commit 6e2476ea4d)
It is all in one line because <http://elinks.cz/authors.html>
is converted from AUTHORS and the conversion script is not
clever enough to know which newlines should be kept.
If the licence of ELinks is ever changed to "GNU GPL v3 or later",
I'll probably stop hacking it, but I don't want to legally forbid
such a change.
Make it clear that modified versions may also be distributed.
I am the sole copyright holder for these ELinks files
so I can replace the licence like this.
This version of the licence is used in bind-9.5.0-P2.tar.gz.
Wikipedia claims ISC made the change in July 2007 after a request
from the FSF.
The relicensing permission was in message
<200808271214.22088.kdudka@redhat.com> posted to elinks-dev on
2008-08-27. COPYING suggests such things should be added to AUTHORS.
The dependency table was originally in "fixed" format, so that
AsciiDoc took a fixed number of characters to the first column
and the rest to the second column. However, "OpenSSL or GNU TLS
or nss_compat_ossl" will not fit in that number of characters,
and backslash line continuation will not help because AsciiDoc
apparently parses that before it counts the characters.
I could widen the column in the AsciiDoc source but switching
to "dsv" format seems prettier.
test/optgroup.html was added on 2004-04-17 with no comment about licensing.
I contacted the author via <http://iccl.fi/feedback.cgi?id=mail>, asking
for a licence. The author noted that a developer of ELinks had originally
asked on the #debian.fi channel whether the file could be used, and he had
allowed it then. That permission grant had not been recorded in the source
tree though, and it is not clear whether modification had been allowed.
Anyway, the author now explicitly grants us the GNU free documentation
licence on this file, and is willing to consider other licences.
never_for_this_site(form) did return remember_form(form).
In ELinks 0.11.0, both functions returned int, so this was OK.
In commit 2b7788614f however, the
functions were changed to return void, as required by msg_box().
GCC still accepted the return statement but Sun Studio 11 did not.
On 2008-09-05, it was reported to elinks-dev that ELinks hits an
internal error (bad alloc_header) when given a specific HTML file.
On 2008-09-09, out-of-range values of document->comb_x and
document->comb_y were noted as the cause of memory corruption.
Update those variables when splitting, aligning, or justifying a line.
Add many assertions to detect the bug if it occurs again.
Previously, the character at (document.comb_x, document.comb_y) was
accessed via the POS macro, which adds part.box.x and part.box.y to
the coordinates. However, if document.comb and document.y are set
at the end of one part and read at the beginning of another, then
the struct screen_char used by the original part should be updated,
even though the new part has a different box. Change comb_{x,y} to
be relative to the document, rather than to the box of a single part.
The GNU Hurd has a bug that can make select() report an exception in a
pipe even though none has actually occurred. The typical result is
that ELinks closes the pipe through which it internally passes all
input events, such as keypresses. It then no longer reacts to what
the user is trying to do.
Work around the Hurd bug by making set_handlers() check whether the
file descriptor refers to a pipe, and if so, pretend the caller did
not provide any handler for exceptions. This is a minimal change that
avoids slowing down the select() loop itself and does not require
careful analysis of the callers to statically find out which file
descriptors might refer to pipes. The extra stat() calls may slow
ELinks down somewhat, but anyway it'll work better than it did without
the patch, and if the Hurd bug is ever fixed, we can remove the
workaround at that time.
Previously, spidermonkey_get_interpreter() and init_smjs() each called
JS_SetErrorReporter on the JSContexts they created. However,
JS_SetErrorReporter actually sets the error reporter of the JSRuntime
associated with the JSContext, and all of our JSContexts use the same
JSRuntime nowadays, so only the error_reporter() of
src/ecmascript/spidermonkey.c was left installed. Because this
error_reporter() asserts that JS_GetContextPrivate(ctx) returns a
non-NULL pointer, and init_smjs() does not set a private pointer for
smjs_ctx, any error in smjs_ctx could cause an assertion failure, at
least in principle.
Fix this by making spidermonkey_runtime_addref() install a shared
error_reporter() when it creates the JSRuntime and the first JSContext.
The shared error_reporter() then checks the JSContext and calls the
appropriate function.
The two error reporters are quite similar with each other. In the
future, we could move the common code into shared functions. I'm not
doing that yet though, because fixing the bug doesn't require it.
make_bittorrent_peer_connection() used to construct a struct uri on
the stack. This was hacky but worked nicely because the struct uri
was not really accessed after make_connection() returned. However,
since commit a83ff1f565, the struct uri
is also needed when the connection is being closed. Valgrind shows:
Invalid read of size 2
at 0x8100764: get_blacklist_entry (blacklist.c:33)
by 0x8100985: del_blacklist_entry (blacklist.c:64)
by 0x80DA579: complete_connect_socket (socket.c:448)
by 0x80DA84A: connected (socket.c:513)
by 0x80D0DDF: select_loop (select.c:297)
by 0x80D00C6: main (main.c:353)
Address 0xBEC3BFAE is just below the stack ptr. To suppress, use: --workaround-gcc296-bugs=yes
To fix this, allocate the struct uri on the heap instead, by
constructing a string and giving that to get_uri(). This string
cannot use the "bittorrent" URI scheme because parse_uri() does not
recognize the host and port fields in that. (The "bittorrent" scheme
has protocol_backend.free_syntax = 1 in order to support strings like
"bittorrent:http://beta.legaltorrents.com/get/159-noisome-beasts".)
Instead, define a new "bittorrent-peer" URI scheme for this purpose.
If the user attempts to use this URI scheme, its handler aborts the
connection with an error; but when make_bittorrent_peer_connection()
uses a bittorrent-peer URI, the handler is not called.
This change also lets get_uri() set the ipv6 flag if peer_info->ip is
an IPv6 address literal.
Reported by Witold Filipczyk.
Introduce static int interpreter_count in src/ecmascript/ecmascript.c.
Maintain interpreter_count in ecmascript_get_interpreter and
ecmascript_put_interpreter.
Introduce ecmascript_get_interpreter_count.
Display the number of ECMAScript interpreters that have been allocated
for documents in the Resources dialog box.
Separate the formatting of unparsed lines from ftp_process_dirlist()
to a new function ftp_add_unparsed_line(). Check for all possible
out-of-memory errors. Encode HTML metacharacters as entity references
and document how charsets are handled FTP directory listings.
Add a NEWS entry.
With Sun Studio 11 on Solaris 9, we get "cc: Warning: illegal option
-dynamic"; then, cc proceeds anyway, but the option can prevent the
linker from finding the libraries listed in -l operands. To detect
this, move the -rdynamic check in configure.in down to a place where
the libraries have already been added to $LDFLAGS. So if -rdynamic
interferes with the search for libraries, ELinks won't use it.
Merely moving the test would also change the location of -rdynamic in
$LDFLAGS. Counteract that by making the test add -rdynamic to the
beginning of $LDFLAGS, rather than to the end. This may make the test
more reliable on Solaris.
ELinks used to call the MD5 code in libgnutls-openssl, part of
GNUTLS-EXTRA, which was licensed under GNU GPL version 2 or later.
In GnuTLS 2.2.0 however, the license of GNUTLS-EXTRA has been changed
to GNU GPL version 3 or later. This is no longer compatible with
GNU GPL version 2 as used in the current ELinks, because GPLv2 clause
2. b) requires the whole work to be licensed under GPLv2, and GPLv3
does not allow that.
If anyone is still using a pre-2.2 GnuTLS, he or she can tweak
configure.in to check the version or just assume it's old enough.
There is not much reason to do so though, as including the MD5 code
in ELinks seems to cost only about 4 kilobytes on i686.
(cherry picked from commit 9ca0182ec6)
cache_entry.id => cache_entry.cache_id
document.id => document.cache_id
ecmascript_interpreter.onload_snippets_owner => .onload_snippets_cache_id
This is a combination of:
commit 232c07aa7f
bug 1009: id variables renamed, added document_id to the document.
commit 6007043458bf8f14abfc18b9db60785bdc0279f6
Revert addition of document.document_id
fsp_open_session() has a bug where it does not set errno if getaddrinfo fails.
Before the bug 1013 fix, this caused an assertion failure.
After the bug 1013 fix, this caused a "Success" error message.
Now it instead causes "FSP server not found".
Replace almost all uses of enum connection_state with struct
connection_status. This removes the assumption that errno values used
by the system are between 0 and 100000. The GNU Hurd uses values like
ENOENT = 0x40000002 and EMIG_SERVER_DIED = -308.
This commit is derived from my attachments 450 and 467 to bug 1013.
In the past, we have hit annoying incompatibilities when
people have attempted to rebuild ELinks documentation
with new versions of AsciiDoc:
http://bugzilla.elinks.cz/show_bug.cgi?id=989http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=491820
So now we bundle the configuration files from a known good version
to let ELinks documentation be built with them regardless of which
version has been installed on the machine.
Add new routine compute_redraw_interval, which returns the appropriate
interval in milliseconds for updating the LED panel, namely 100ms if there
are any downloads, 1000 if the clock is enabled (with a TODO noted to check
whether the date format includes seconds), or 0 otherwise to indicate that
the LED paanel need not be updated
Use the new compute_redraw_interval routine in draw_leds and redraw_leds.
This fixes bug 973, "LED indicators wake system up every 100ms".
On Mac OS X 10.5.4, <net/if.h> does not #include <sys/socket.h> but
uses struct sockaddr defined there. Autoconf 2.61 generates a
configure script that warns if the header can be preprocessed but not
compiled. The Autoconf manual cautions that future versions of
Autoconf will treat the file as missing in this case. To let ELinks
detect <net/if.h> even with a future Autoconf, make the test program
#include <sys/socket.h> before <net/if.h>.
It seems GnuTLS is not as good at negotiating a supported protocol as
OpenSSL is. ELinks tries to work around that by retrying with a
different protocol if the SSL library reports an error. However,
ELinks must not automatically retry POST requests where some data may
have already reached the server; POST is not a safe method in HTTP.
So instead, collect the name of the TLS-incapable server in a blacklist
when ELinks e.g. loads an HTML form from it; the actual POST can then
immediately use the protocol that worked.
It's a bit ugly that src/network/socket.c now uses
protocol/http/blacklist.h. It might be better to move the blacklist
files out of the http directory, and perhaps merge them with the
BitTorrent blacklisting code.
