http_negotiate_parse_data passed &token->length as the int *outlen
parameter of base64_decode_bin, which stores an int at that location.
However, gss_buffer_desc::length is size_t in all implementations that
I checked: MIT Kerberos Version 5 Release 1.10, libgssglue 0.4, and
GNU GSS 1.0.2. This mismatch could cause the build to fail:
.../src/protocol/http/http_negotiate.c: In function ‘http_negotiate_parse_data’:
.../src/protocol/http/http_negotiate.c:173:2: error: passing argument 3 of ‘base64_decode_bin’ from incompatible pointer type [-Werror]
In file included from .../src/protocol/http/http_negotiate.c:30:0:
.../src/util/base64.h:8:16: note: expected ‘int *’ but argument is of type ‘size_t *’
On 64-bit big-endian hosts, it might also cause the GSSAPI
implementation to read too much data from memory and disclose it to
some network server, or crash ELinks.
I had already done this to my other scripts on 2008-09-28 (commit
c67885d880) but missed Unicode/gen-case.
Update it, and list it in COPYING.
(Although Unicode/gen-case is part of the source tree, this version of
ELinks does not use that file for anything.)
The HTML parser decoded SGML entity references and numeric character
references in the following attributes, and then the renderer did the
same again:
link/@title
link/@hreflang
link/@type
link/@media
img/@alt
area/@alt
input[@type="image"]/@alt
input[@type="image"]/@name
input[@type="button"]/@value
The result was that e.g. title="A" displayed as "A"
even though it was supposed to display as "A".
Fix by making the HTML parser tell the renderer that the entities have
already been decoded.
(cherry picked from commit 715571a5d6)
If ELinks was configured --without-x, the build failed at
xprop_to_string:
.../src/osdep/osdep.c:500: error: expected ‘)’ before ‘*’ token
Fix by moving xprop_to_string into the #ifdef HAVE_X11 section.
Reported by Thomas Adam.
(cherry picked from commit e3cb8d6a77)
If Xutf8TextPropertyToTextList or XmbTextPropertyToTextList returns a
positive number, that means some characters were unconvertible and
have been replaced with XDefaultString(). Use the resulting string
even in that case, as if the function had returned Success.
The previous version ignored the string and didn't even free it.
(cherry picked from commit d2cf6912b9)
gnome-terminal 2.30.2 expects UTF-8 in the "OSC Ps ; Pt BEL" sequence
that sets the window title. However, XGetWMName typically returns the
title in "STRING" (Latin-1) or "COMPOUND_TEXT" (escape sequences)
encoding. Recode the title to restore it correctly. This helps
especially in the fi_FI.UTF-8 locale, where gnome-terminal has "Pääte"
as the default title.
Related to bugs 885 and 336.
(cherry picked from commit d1245c73a8)
Make u2cp_() map code points U+0080 to U+009F via strange_chars[] even
if the target codepage is UTF-8. This helps with buggy web pages that
use ’ when they mean ’. This change does not affect how
ELinks decodes raw bytes 0x80 to 0x9F in HTML.
u2cp_() is used only via the u2cp and u2cp_no_nbsp macros.
Possible side effects of this change at each use of these macros:
* get_translation_table(): Not affected because it does not call u2cp
if the target codepage is UTF-8.
* get_entity_string(): Numeric character references are affected, as intended.
Character entity references are not affected because entities[]
does not define any entities in the U+0080...U+009F range.
* kbd_field(), term_send_ucs(), field_op(): Affected. It is no longer
possible to enter code points U+0080...U+009F from the terminal.
This should not be a problem in practice because those would be
control characters anyway and should therefore be filtered by the
slave process (which doesn't yet recognize them; bug 777).
unregister_options() requires as a sentinel an instance of struct
option where option.name is NULL. However, the NULL_OPTION_INFO macro
used for these sentinels actually initializes a struct option_init
instead. Make register_options() overwrite the NULL_OPTION_INFO with
a sentinel in the correct format. This probably makes a difference
only on platforms where null pointers don't have all bits zero.
l_set_option() was passing the address of an int to
option_types[OPT_INT].set and option_types[OPT_BOOL].set.
That looks correct but is not: both function pointers
point to num_set(), which actually reads *(long *) str.
Change l_set_option() to pass the address of a long instead,
and add comments about this dependency.
(cherry picked from commit 8766e3829f)
ELinks already allowed '+' and '*' in the names of options when
reading a configuration file. The option manager however didn't
let the user add such options. Allow the characters there too.
These characters are needed especially in the mime.type tree,
where '*' is used as a replacement for '.'. For example:
set mime.type.audio.prs*sid = "sid"
set mime.type.application.atom+xml = "atom"
This commit changes one gettextised string.
The 2009-08-15 fix for bug 1083 made read_encoded() return -1 of EOF,
like decompress_data() expects. Unfortunately, read_file() too calls
read_encoded(), and it treated the -1 as an error and reporter
whatever error had been left in errno. This made it impossible to
display local files, compressed or not.
Apparently then, read_encoded() needs to distinguish between decoded
bytes, EOF, EAGAIN, and true errors. Make it return an enum.
INIT_OPTION used to initialize union option_value at compile time by
casting the default value to LIST_OF(struct option) *, which is the
type of the first member. On sparc64 and other big-endian systems
where sizeof(int) < sizeof(struct list_head *), this tended to leave
option->value.number as zero, thus messing up OPT_INT and OPT_BOOL
at least. OPT_LONG however tended to work right.
This would be easy to fix with C99 designated initializers,
but doc/hacking.txt says ELinks must be kept C89 compatible.
Another solution would be to make register_options() read the
value from option->value.tree (the first member), cast it back
to the right type, and write it to the appropriate member;
but that would still require somewhat dubious conversions
between integers, data pointers, and function pointers.
So here's a rather more invasive solution. Add struct option_init,
which is somewhat similar to struct option but has non-overlapping
members for different types of values, to ensure nothing is lost
in compile-time conversions. Move unsigned char *path from struct
option_info to struct option_init, and replace struct option_info
with a union that contains struct option_init and struct option.
Now, this union can be initialized with no portability problems,
and register_options() then moves the values from struct option_init
to their final places in struct option.
In my x86 ELinks build with plenty of options configured in, this
change bloated the text section by 340 bytes but compressed the data
section by 2784 bytes, presumably because union option_info is a
pointer smaller than struct option_info was.
decompress_data() supposed that read_encoded() would return a positive
number if it decompressed something, 0 if no data is available yet but
may be later, or -1 if no more data will be available. However,
several backends actually returned 0 if they had seen an EOF marker or
an error in the stream, causing decompress_data() to keep calling
them. Make them return -1 in this situation.
http://buildd.debian-ports.org/fetch.php?pkg=elinks&arch=hurd-i386&ver=0.12%7Epre5-1&stamp=1247378205&file=log&as=raw
shows two unsightly errors, apparently triggered by TERM=unknown:
tput: unknown terminal "unknown"
/build/buildd/elinks-0.12~pre5/configure: line 25480: test: -ge: unary operator expected
Hide the first of them by redirecting to /dev/null,
and avoid the second one by substituting "0" if tput fails.
I'm not sure why `which tput` is used instead of plain tput,
but it doesn't seem to be causing any problems in practice.
(There are problems in theory if tput is not installed,
tput is found in a directory whose name contains spaces,
or tput is a shell function or builtin.)
Fix this GCC 3.4.6 warning, which becomes an error
if configure --enable-debug adds -Werror to CFLAGS:
[CC] src/document/css/apply.o
In file included from /home/Kalle/src/elinks-0.12/src/document/html/internal.h:6,
from /home/Kalle/src/elinks-0.12/src/document/css/apply.c:35:
/home/Kalle/src/elinks-0.12/src/document/html/parser.h:149: warning: parameter has incomplete type
In file included from /home/Kalle/src/elinks-0.12/src/document/css/apply.c:35:
/home/Kalle/src/elinks-0.12/src/document/html/internal.h:125: warning: parameter has incomplete type
Even without this warning, "enum html_special_type;"
would not be standard C89.
location_goto() was defined identically in
src/ecmascript/spidermonkey/location.c and
src/ecmascript/see/location.c, and declared in four other files.
Merge both definitions to the new src/ecmascript/location-goto.c
and all declarations to src/ecmascript/ecmascript.h.
Call get_window_title() at most once per terminal, even if it fails.
If ELinks is configured with X11 support, get_window_title() calls
XOpenDisplay(), which can sleep if $DISPLAY does not respond.
This delay was previously incurred every time ELinks changed the
title. Besides, if ELinks had already changed the title, setting
ditrm->orig_title = get_window_title(); did not make sense anyway.
Bring the code closer to ELinks 0.13.GIT commit
71ccbe0f8d made on 2007-11-07.
Don't define a separate download_flags_T typedef though,
because Doxygen generates nicer links if the enum is used directly.
There are two identical tp_show_header() functions: one in
src/session/download.c and one in src/protocol/bittorrent/dialogs.c.
Neither is declared in any header, but the latter was not static.
If msg_box() runs out of memory, it returns NULL. In this case,
the done_handler_T callbacks of the buttons will not be called. So
lookup_unique_name() must instead free the struct lun_hop on its own.
Remove the stracpy(ofile) call that could never be executed.
This removes the need to handle errors from that call,
and makes it clear that lun_hop->file need not be separately freed
if an error occurs in lookup_unique_name().
