The previous scheme incorrectly accepted 0xC1 0x80 as U+0040.
That could have been fixed by tweaking the loop, but the constant
array is surely easier to verify.
- Include arpa/inet.h to get hton* ntoh* functions.
- Use socklen_t instead of int.
- Try to define PF_INET to AF_INET if it doesn't exist.
Reported-by: Andy Tanenbaum <ast@cs.vu.nl>
In the previous version, invalid UTF-8 from a terminal caused
UCS_NO_CHAR (0xFFFFFFFD) to be stored in a term_event_key_T, resulting
in -3 which was then incidentally treated as an unassigned special key.
Now, invalid UTF-8 is instead mapped to UCS_REPLACEMENT_CHARACTER
and treated as a character. The fact that handle_interlink_event
calls term_send_ucs when it receives invalid UTF-8 makes it pretty
clear that this is how it was intended.
src/viewer/text/link.c (not changed in this commit) already referred
to UCS_REPLACEMENT_CHARACTER in a comment even though it was not
previously defined.
The message appears when the user has selected e.g. "Main mapping"
rather than an action inside it. Because the main mapping is a keymap,
ELinks must not tell the user to select a keymap.
whether we ought to add the conn->progress->start to
the conn->est_length. Currently displaying resuming works correctly with
ftp.task.gda.pl and ftp.pld-linux.org.
Restring the spaghetti in dump_to_file to fix a bug that was introduced
in commit 2a6125e3d0 whereby when
document.dump.codepage != "utf-8", the document itself was not output,
only the references list.
Decrement term->current_tab before calling delete_window() instead of after
deleting all backgrounded tabs, so get_tab_by_number() will see a
consistent value.
It cannot be restricted just to characters that have passed
check_kbd_label_key(), because hotkeys in strings received from
gettext must also be processed with it, and there we don't have
a struct term_event for check_kbd_label_key().
This causes the documented-slow cp2u() to be called in a loop, which
fortunately doesn't have very many iterations. If this is too slow,
then cp2u() can be rewritten, or the hotkeys can be cached in struct
widget or struct widget_data.
Note that check_kbd_label_key() does not yet allow non-ASCII
characters when CONFIG_UTF_8 is defined. Before they are allowed,
menu.c should also be updated.
According to Jonas Fonseca, if init_string(&canonical) fails, then it
anyway sets canonical.source = NULL and makes done_string(&canonical)
safe, even if canonical was previously uninitialized.
Actions can now be bound to e.g. Ctrl-Alt-A. The keybinding code also
supports other combinations of modifiers, like Shift-Ctrl-Up, but the
escape sequence decoder doesn't yet.
Don't let Ctrl-Alt-letter combinations open menus.
The cast is not necessary since we already check the bounds, but by using a
cast here, it hopefully makes it more obvious what the long comment above
is pointing out: namely that we put the value of a signed integer into an
unsigned char.
This fixes a bug: in the previous version, l_bind_key() modified the
buffer whose address lua_tostring() returned, even though that is not
allowed according to Lua documentation <http://www.lua.org/pil/24.2.2.html>.
The change affects the user interface: previously, if the user typed
"ctrl+cokebottle" in the "Add keybinding" dialog box, ELinks would
change the text in the widget to "Ctrl-cokebottle" before complaining
that the keystroke is invalid. Now, it leaves the widget unchanged.
This commit does not yet add const to parameters of parse_keystroke()
and related functions.
Before really_add_keybinding() is called, check_keystroke() calls
parse_keystroke(), which converts the modifier prefix to canonical
form: for example, "alt+f9" becomes "Alt-f9". This commit makes
really_add_keybinding() normally ignore that string and generate a
brand new one, e.g. "Alt-F9" (note the upper-case F), for its
"Keystroke already used" warning. Likewise, " " turns to "Space".
After this commit, it should be possible to change parse_keystroke()
to never write back into its input string.
If really_add_keybinding() cannot generate the string for some reason
(out of memory?), then it will use whatever parse_keystroke() has left
in the buffer. The alternatives would be to omit the keystroke name
from the warning or to reject the keybinding entirely; it isn't clear
what the best solution is here, but the one I implemented is closest
to the previous behaviour.
This requires compiling cp2u() in even without CONFIG_UTF_8.
I also added an is_kbd_character macro to make try_document_key
more resilient to changes in the definition of term_event_key_T.
The previous version used only the low 8 bits of the key code.
This one arranges for the whole key to be rejected if it's not ASCII.
Perhaps the modifiers should be checked too, but I'm not changing that now.
In revision 1.15 of dns.c (as it was called way back then), pasky
backported a fix from Links 0.97pre2 to try gethostbyaddr before
trying gethostbyname for DNS lookups:
MacOS address resolution fix (Aldy Hernandez) (from 0.97pre2)
However, that fix introduced a bug, because it was calling gethostbyaddr
on all addresses, not just IP addresses. Mikulas fixed that bug in Links
0.98:
Do not call gethostbyaddr when name is not ip address (it should avoid
some useless nameserver queries)'
This fix was never backported to ELinks. Until today.
This commit is functionally the same as the fix in Links 0.98, plus it uses
inet_aton for great correctness!
This fixes a bug reported in #elinks by tnks, whereby lookups for
yubnub.org resulted in 121.117.98.110 == 0x7975626E == 'y', 'u', 'b', 'n'.
I believe that it also fixes bug 691 (which is already closed with a
workaround).
To reproduce before this patch:
- Run ELinks with an 80x25 terminal.
- Set document.browse.forms.confirm_submit = 1.
- Go to <http://bugzilla.elinks.cz/query.cgi>.
- Click the [ Search ] submit button.
- ELinks asks "Do you want to post form data to URL".
Each line of the URL begins at the horizontal center of the dialog,
and bleeds outside the right border of the dialog. Also, the
[ Yes ] and [ No ] buttons appear to float below the dialog.
Previously, ELinks used to silently discard the Alt modifier from
Alt- keystrokes when UTF-8 I/O was enabled. Now, separate actions
can be bound to and Alt-.
However, if CONFIG_UTF_8 is defined, then actions cannot be bound to
non-ASCII characters, regardless of modifiers. This is because the
code that handles names of keystrokes assumes a character can only be
a single byte. This commit does not change that.
Form fields and BFU text-input widgets then convert from UCS-4 to UTF-8.
If not all UTF-8 bytes fit, they don't insert anything. Thus it is no
longer possible to get invalid UTF-8 by hitting the length limit.
It is unclear to me which charset is supposed to be used for strings
in internal buffers. I made BFU insert UTF-8 whenever CONFIG_UTF_8,
but form fields use the charset of the terminal; that may have to be
changed.
As a side effect, this change should solve bug 782, because
term_send_ucs no longer encodes in UTF-8 if CONFIG_UTF_8 is defined.
I think the UTF-8 and codepage encoding calls I added are safe, too.
A similar bug may still surface somewhere else, but 782 could be
closed for now.
This change also lays the foundation for binding actions to non-ASCII
keys, but the keystroke name parser doesn't yet support that.
The CONFIG_UTF_8 mode does not currently support non-ASCII characters
in hot keys, either.
There is no need to check whether ev->ev == EVENT_KBD;
if decode_terminal_escape_sequence called
decode_terminal_mouse_escape_sequence, then the former neither modified
kbd.key nor passed &kbd to the latter, so kbd.key remains KBD_UNDEF.
If ev->ev was not checked, then it should not be trusted either.
So reinitialize the whole *ev if a keyboard event was indeed found.
For instance, if Ctrl-F1 were pressed and src/terminal/kbd.c supported it,
then toupper(KBD_F1) would be called, resulting in undefined behaviour.
src/terminal/kbd.c does not support such combinations yet, but it is
safest to fix the bug already.
... since the latter is for printing int64_T and we don't check for that and
we use PRId64 only for printing values having the off_t types.
Besides off_t has it's own ELinks specific defaults so it should be safer
to use an internal format string. If off_t is 8 bytes use "lld" else use
"ld".
Reported-by: Andy Tanenbaum <ast@cs.vu.nl>
Drop some code for superscript and subscript handling that was deleted
in commit 65016cdca4, then added back
with the UTF-8 merge in commit 2a6125e3d0,
and then disabled in commit 1b653b9765.
Also list the capnames with which the escape sequences could be
read from Terminfo, and the ECMA-48 interpretations of the bytes
(parenthesized if they seem unrelated to the keys). This is in
preparation for fixing bug 96.
decode_terminal_escape_sequence() used to handle both, but
there is now a separate decode_terminal_application_key()
for ESC O. I have not yet edited decode_terminal_escape_sequence();
there may be dead code in it.
If there is e.g. ESC [ in the input buffer, combine that to Alt-[.
Check the first character too; don't blindly assume it is ESC, as
it can be NUL as well. Note this means you can no longer activate
the main menu by pressing Ctrl-@ (or Ctrl-Space on some terminals).
Otherwise, the timeout could cause ELinks to resume reading from
the terminal device while another process is still using it.
This actually happened in a test.
On entry to some functions that could resume reading from the device,
assert that the terminal has not been blocked.
To reproduce the bug before this patch:
Enable CONFIG_UTF_8, UTF-8 I/O, and UTF-8 charset.
Go to www.google.com and type "abc" in the text input field.
Then press Left. The cursor jumps to "a" when it should go to "c".
to encode it using utf_8_to_unicode. If every unicode_val_T value
could be a result of that function then one must add out param
to the utf_8_to_unicode signaling 'true' UCS_NO_CHAR.
GPM is awful, tho', and this is an ugly hack. Why can't it just report
buttons like everyone else?
Another nice thing is that we don't seem to get the wheel events more
frequently than once in a second or so. Therefore it will work properly
only when you scroll slooooowly. :^)
When the user presses enter during a text type-ahead search, simply cancel
the search without additionally following the current link. Link type-ahead
searching still will follow the active link on enter.
using magic ;)
Now in resume mode connection is always interrupted
and resumed. Even when all file is downloaded from beginning
conection will be resumed from old end of file. Feel free to fix it.
In protocol/common.c length of string is known, so pass it
instead of -1 to encode_uri_string.
Introduced encode_win32_uri_string, because there were problems
with : and \ in base href.
Don't call clear_dialog, which sets the focus to the listbox. Neither the
button widget nor the listbox widget has a clear callback, and the only
other thing that clear_dialog does is focus the first widget and redraw, so
call redraw_dialog instead.
Thanks to Kalle Olavi Niemitalo for noticing the issue.
Make move_up and move_down return no value. Instead, save the old y value
and compare it to the new after calling move_up or move_down in
move_page_up or move_page_down, respectively.
This fixes a bug where if given a prefix, if that prefix specified a number
of pages greater than move-page-up actually scrolled, there would be no
screen update, because the last call to move_up would return FRAME_EVENT_OK
which would be returned from move_page_up, even tho move_page_up would have
previously returned FRAME_EVENT_REFRESH.
Don't try the key as an accesskey if a menu was opened, whether it was just
the main menu or whether it was a submenu of the menu menu (we would try
the key as an accesskey in the latter case).
In send_kbd_event, replace the KBD_MOD_ALT modifier when trying the key as
an accesskey rather than when we don't.
Commit 3ce3f01f30 introduced a bug whereby
if a tab set the current position in the line to or greater than the number
of bytes remaining in the source, the line was split after the tab.
Delete the FREE_LIST flag from mi_no_ext so that free_menu_items doesn't
try to free static memory.
This fixes a crash reproducible by deleting every entry under Setup -> File
extensions, opening said menu, and then closing said menu.
Simplify do_move_bookmarks (from a readability perspective) by replacing
the destb and desti parameters with a new insert_as_child flag. Inspired
by Kalle Olavi Niemitalo.
do_move_bookmark was only updating the selection in the bookmarks manager
window in which the Move button was pressed. Now all windows are updated.
This patch also prevents a crash when the first item that was displayed
in a box was the last child of a folder and was being moved (the comment
removed in this patch was incorrect in assuming that bm->box->next must
be valid because it neglected to account for non-root children).
This change required that I move the definition of struct
hierbox_dialog_list_item from src/bfu/hierbox.c to src/bfu/hierbox.h.
Thanks to Kalle Olavi Niemitalo for finding both the update problem
and the crash.
Before, when one assigned a value other than a function to
elinks.preformat_html, elinks.goto_url_hook, or elinks.follow_url_hook,
an assertion failure was triggered (and probably worse happened in
non-DEBUG builds). Now just fail silently.
Increment conn->from by the length of the data so that when
abort_connection calls normalize_cache_entry, it doesn't truncate the
cache entry to 0 length.
It is similar to lua's execute and let's you run a command line in
'non-blocking' mode. Example:
elinks.keymaps.main["F"] = function () {
elinks.execute("firefox " + elinks.location);
};
All cookies are now constructed with the new function init_cookie.
Requested by Miciah Dashiel Butler Masters.
This also fixes a bug where the "Add cookie" button left cookie->path == NULL,
causing a crash later when deciding whether to send the cookie to the server.
Mostly this makes set_cookie more complex, as it now distinguishes
between HEADER_PARAM_NOT_FOUND and HEADER_PARAM_OUT_OF_MEMORY, and kills the cookie in the
latter case. However, the cookie->secure check became simpler.
Its return value is now an enum that lets callers know whether an
error occurred. However, this commit changes the callers only
minimally, so they do not yet check the return value.
Previously, it was a pointer to a separately allocated struct attributes,
of which only the LIST_HEAD members were actually used.
All init_list calls now operate on struct list_head.
This change may also reduce memory usage for reading bookmarks.
Also, add the server in the done_handler_T of the "OK" button,
rather than in the widget_handler_T of the "Server name" field
(which check_dialog calls). This change implies that the dialog
is closed even if an error occurs while adding the server, but
that seems to be customary elsewhere in ELinks too.
src/cookies/dialogs.c (set_server): Set up a struct cookie and pass it
to accept_cookie, instead of calling set_cookie which might ask the user
to confirm. (This change also removes a dubious use of struct uri.)
src/cookies/cookies.c (get_cookie_server): No longer static.
src/cookies/cookies.h (get_cookie_server): Declare.
With regular comments in the definition of the structure itself,
and with xgettext:c-format comments in constants of that type,
if xgettext would otherwise guess wrong; so that translators
will know they'll have to double any percent signs they add.
I didn't regenerate PO files, though.
In html_subscript, html_subscript_close, html_superscript, html_quote, and
html_quote_close, use put_chrs instead of html_context->put_chars_f.
Element handlers should use put_chrs so that it can correctly handle
whitespace and stuff.
Instead of saving the old link colours when selecting a link and using that
to restore them when unselecting it, just copy the data from the document.
- Eliminate struct link_bg and the .link_bg and .link_bg_n members
of struct document_view.
- Eliminate the free_link routine and don't call it from draw_doc,
clear_link, or detach_formatted.
- Add a .old_current_link member to struct view_state and initialise it in
init_vs.
- Don't save link_bg in draw_current_link.
- Rewrite clear_link to use the document data instead of link_bg.
- Modify init_link_drawing not to allocate ling_bg and to return a pointer
to a static variable for the template character.
Introduce html_subscript_close callback. Draw opening and closing brackets
and carets for subscript and superscript text directly in the element
handlers rather than performing weirdness in the renderer. This both
improves readability and fixes bug 284, misplaced brackets with subscripts.
Add close callbacks html_html_close, html_style_close, and
html_xmp_close. end_element now calls the element close callback instead
of performing special handling for certain tags.
src/bfu/menu.c (scroll_menu): Let neither menu->selected nor pos
become -2.
src/bfu/menu.c (menu_mouse_handler): Call set_menu_selection directly
rather than via scroll_menu, as sel is already known to be selectable.
(Not required for fixing the bug.)
src/bfu/menu.c (menu_search_handler): Break infinite loops also if
menu->selected is -1 initially.
src/bfu/menu.c (menu_handler): Instead of tweaking menu->selected
directly, let scroll_menu do it.
This fixes two bugs:
1. Pressing F9 did not make the main menu visible, but then pressing
e.g. Right made it visible.
2. Pressing F9 and then Down displayed the first submenu (File) at the
wrong position on the screen.
src/terminal/tab.c (get_tab_by_number): Assert that the returned
struct window * actually points to a struct window.
src/terminal/tab.c (move_current_tab): Keep tabs contiguous in the
stack of windows. Obey "ui.tabs.wraparound".
src/terminal/terminal.h (struct terminal): Documented that tabs can
move in the stack.
src/terminal/window.c (assert_window_stacking): Check that the main
menu isn't between tabs.
This fixes the bug that tabs opened with -remote used to hide existing
dialogs, which then became unusable until the new tabs were closed.
src/terminal/tab.c (init_tab): Put the new tab immediately above
existing ones, or if it's the first one, then at the bottom of the stack.
Added assertions.
src/terminal/terminal.h (struct terminal): Redocumented the stacking
order of windows.
src/terminal/window.c [CONFIG_DEBUG] (assert_window_stacking): New function.
src/terminal/window.h (assert_window_stacking): New function or no-op macro.
src/bfu/README: This new file currently contains a diagram of how the
various struct types of src/bfu/ and src/terminal/ relate to each
other. More documentation may be added later, although if it is
specific to a particular structure, then it should probably go in the
corresponding header file so that people will remember to update it.
Including double-width glyph support.
Note: textarea is now drawn with blank collumn at end of lines. It seems
that this is more intuitive for users. It behaves similar as textareas in
graphical interfaces. I hope it will hold your interest.
+----+ +----+ +----+
|aaA | [right] |aaa_| [right] |aaa |
|aaa | |aaa | |Aaa |
|bb | |bb | |bb |
+----+ +----+ +----+
+----+ +----+ +----+
|Aaa | [end] |aaa_| [c] |aaa |
|aaa | |aaa | |Caa |
|bb | |bb | |abb |
+----+ +----+ +----+
A, _, C - cursor positions.
[right] - right arrow
[end] - ACT_END (End button)
[c] - Letter c.
Now this code:
<textarea rows="3" cols="3">aaaaaabb</textarea>
represents textarea with 3x3 positions for chars.
Before this texteare behaved some kind of weirdly. That code above was
rendered like this:
+----+
|aaaa|
|aabb|
|_ |
+----+
at least under Linux. I didn't test network part of external handlers.
Files in tmpdir must be delete somehow. Maybe make a list of files
to unlink and delete them while quitting ELinks.
copiousoutput III. It works only for local files, but you must remove
| /usr/bin/less
yourself and allow reading special files, and set default mime type
as text/plain for convenience.
Do not add the label for the text field to the memory list (the list of
allocations that should be freed with the dialogue).
Thanks to Kalle Olavi Niemitalo for reporting this bug.
Items can be accessed via elinks.globhist["http://www.foo.net/"]. The
resulting object has title, url, and last_visit properties. last_visit
is the seconds since the epoch. See comments in the code for why I don't
provide a nicer interface for last_visit.
Introduce a new list, global_history_reap_list, and a new routine,
reap_deleted_globhist_items, which deletes unused items from that
list. In done_global_history_item, do not free the item; just move it
to that list. Call global_history_reap_list in add_global_history_item
and free_global_history.
Note: there are ugly bug (feature?) - when there isn't enought room for
whole double-width char two double-chars are displayed. Can be seen on
table with double-width chars reduced as much as possible.
Note: there is ugly hack in ACT_EDIT_BACKSPACE where last byte of UTF-8
character is removed and then moving to left until complete UTF-8
character is found.
Thus, each caller must now choose the accelerator key and declare the
accelerator contexts (i.e. menus) to which it may add the command.
Also, use only one context for tab_menu.
These changes fix the following bugs in accelerator conflict detection:
* "~Pass frame URI to external command" may be displayed together
with "Pass tab URI to e~xternal command", but that was not
declared.
* "Pass link URI to e~xternal command" was declared as being in
the tab menu, but it is actually displayed in the link menu.
Create and immediately destroy a context when initialising the SMJS
document scripting module, because SMJS crashes on exit if there has
been no context created since it started, which is the case if one does
not load any documents.
then dump_to_file_256 is defined in dump.c but not used.
If configure --enable-debug was used, then gcc warns about
the unused function, and the warning stops the build.
2. The description of document.dump.color_mode ends with a
newline, provoking a runtime warning from check_description
in src/config/options.c.
3. options.inc has preprocessor directives inside macro arguments.
That is not portable C. xgettext (GNU gettext-tools) 0.14.3 is
not smart enough to figure out the possible combinations, and
copies an incorrect string to elinks.pot.
Don't replace UTF-8 bytes with '*'. Probably there is need to do better
check what will be displayed.
Also get_current_link_title is no longer pretty and trivial. (o:
With UTF-8 support in terminal enabled it is possible to use double-width
UTF-8 strings as margins of buttons. Although they are displayed wrong when
UTF-8 support in terminal is disabled.
Preparation for using struct terminal in formating functions.
By now distinguish between formating widgets and formating widgets with
displaying was done with term == NULL and term != NULL. I hope I'am
not wrong.
Replace the ECMAScript module initialisation and de-initialisation
routines that wrapped the SMJS and SEE module initialisation and
de-initialisation routines by having the module system call the SMJS
and SEE routines its own darned self.
This patch modifies ELinks wrapping behavior slightly.
* The wrap command now toggles line wrapping in HTML mode, as well as
text mode. Note that when the HTML view of a page is wrapped, its
source view is unwrapped, and vice versa.
* Tabs in text-mode lines are now handled correctly.
* Wrapping a line that reaches exactly to the edge of the screen will
no longer produce a blank line in text mode.
* Text within extra-wide table cells is now wrapped to less than the
screen width, to eliminate sideways scrolling.
The last point is only enabled by setting TABLE_LINE_PADDING to a
non-negative number, in the src/setup.h header file, because it is a
significant change of behavior from previous versions.
Use info_box directly instead of via report_scripting_error by way of
alert_smjs_error, thereby avoiding the addition of extra text saying
that an error has occurred.
Introduce the macros before_widgets and foreach_widget_back. Use the
latter in update_all_widgets instead of foreach_widget so that the
widgets are printed in reverse order, which means that any listbox is
drawn last, which allows it to grab the cursor from the selected button
when the dialogue box is initialised or redrawn.
Requested by Kirk Reiser for great usability with screen readers.
With commit 637f1e82e6 ('NET: Merge
change_connection into cancel_download'), cancel_download returns
immediately if the connection is not in a result state, so save some
code by not checking is_in_progress_state before calling cancel_download.
Revert commit 2f0490cb04
('Eval embedded scripts at once') and follow-up commit
997f61bb32 ('Use document_view instead of
view_state. It is safer probably') because the change causes crashes on
numerous pages and just looks wrong.
This is utterly silly and breaks plenty of stuff - Witek, you CAN'T append
slashes to all hosts just because some silly referer needs it, everywhere
else we NEED the hosts without slashes!
This is another follow-up regression fix that made open and save actions in
the WTD-dialog not function correctly when the connection ended before they
were pressed.
Related: 347970988d
First the patch makes sure doc_loading_callback is always called from
tp_display even if the download is in result state. This is often the
case for local files that the user decides to display via the WTD dialog.
Furthermore, improve the adding to globhist part of doc_loading_callback,
so that it works also for downloads in result state where the
download->conn member is NULL. In addition to grabbing the URI from the
connection try also the cache entry if it is set.
Fixes: bug 355 (Documents displayed via WTD aren't added to globhist)
Commit 9cc9db4e24 broke submit buttons
on HTML forms, so that the server no longer knows which button was
actually pressed.
<kahmalo> The bug with forms seems to be that try_submit_given_form (in
src/viewer/text/link.c) is the only function that runs "onsubmit"
scripts, and it does not care which of the submit buttons was
pressed; it calls submit_given_form which submits based on the first
item of the form. [20:57]
<kahmalo> or last, I don't know how the list works.
<kahmalo> try_submit_given_form could get the control via
get_current_link(doc_view) but I suppose it'd be cleaner to provide
that as a parameter. [20:58]
Originally posted as:
<mid:87ek2heebh.fsf@Astalo.kon.iki.fi>
<nntp://news.gmane.org/87ek2heebh.fsf@Astalo.kon.iki.fi>
<http://permalink.gmane.org/gmane.comp.web.links/2745>
This makes move_download not assume that there is a connection attached
when it is called. This is the case pretty often for file:// downloads when
dialogs are involved (query file) and the reason why it explicitly checks
if the connection state is 'in result state'. Anyway, fill the new download
struct with variabled from the old one instead of taking variables from the
connection struct.
This patch also adds some assertions and a few comments.
This includes setting the new priority and adding the download to the list
of connection downloads. If the connection has no downloads set the
PRI_CANCEL priority; get_priority() requires that.
This simplifies unqueuing of downloads and makes it more obvious that
the 'change' being performed is to migrate or replace an old download
handle with a new one.
Take a quick stroll through the unchartered corners of the DOM node data
structures:
- Remove ununsed struct dom_node_id_item.
- Make the document node reference a future struct dom_document.
- Describe ideas for node data, e.g. the entity reference node should use
it for storing the unicode_val_T.
This ensures that the 'writer' process will remove itself when the
main ELinks process for some reason decides to shutdown the connection.
Before the 'writer' process would complete it's task taking up unnecessary
system resources.
This is mostly an issue when fetching big files. Therefore only file
fetching is fixed. FIXME added about also checking return codes for write
associated with directory listing.
Reported-by: zas
It should be included via elinks.h but apparently some other system header
can prevent this somehow on some systems.
Reported-by: Phillip Pi <ant@zimage.com>
It uses mangleme by Michal Zalewski <lcamtuf@coredump.cx> to generate HTML
which is then fed into the sgml-parser program. By default 100 random HTML
documents are tested. But the test script takes the number of documents
to test against as an argument. Useful for torture testing the SGML parser.
It didn't check that both title and title->text was non NULL. In either
case it now passes "No title" string to add_bookmark().
Reported by Neuromancer.
Tested with both:
<bookmark href="empty://title"><title></title><bookmark>
<bookmark href="no://title"></bookmark>
This was cause by the recent change to allocate string during incremental
parsing where the node string was set after insertion. Test for this in the
works.
Fixes: b6b6d3c67e
Simplify commit 8d4f44f2f1, in particular
detecting MIME types for files. It is more consistent to do it the way
it was already done by the session/download code.
Instead, write a NUL byte to stderr when getting FSP files and only set
cache->content_type when the header string is non-empty.
Additionally it also moves close(stderr) after the fsp_error() in the
file handling part of do_fsp() so the error message is shown with the
correct type.
Fixes problems with host or protocol parts not being lowercased. This
triggers an assertion failure when trying to download such links. Reported
by lindi-.
Use enum connection_state instead of int in load_uri,
proxy_uri, get_proxy_worker, and get_proxy_uri. See commit
d18809522e. I hope that satisfies TCC.
This changes init_dom_node_() to take an allocated argument saying whether
to allocate or not. If the value is -1, node->allocated will be set to the
value of node->parent->allocated. This way the value is inherited like we
do it in the menu code. It should be a sane default since we eventually
want not to rely on the 'underlying' source of the document and there will
be less variables to pass around.
When doing incremental rendering we now require the whole thing to be there
and that there is room for two tokens in the scanner token table. This is
necessary because we have to generate both a processing target token and a
processing data token to make life simpler for the parser.
Remove processing instruction data case label from the main parser loop. It
is safer this way since it already assumes that the processing target token
has been stored.
Check whether there are '=' and value tokens before handling them. If there
is any doubt the whole attribute structure is 'pushed back' into the
stream. That way incremental parsing will not add the value as a new
attribute because the name token was handled in the previous parsing run.
It is a loop that parses the same small document with various read sizes.
The sgml-parser program is updated to take --stdin option taking a the read
size as a required parameter.
That is, add the last parts that saves and resumes previous incomplete
parsing states. Now the parsing stack push handler checks if the parent has
a resume flag set. When set, the incomplete fragment to resume is restored
and the new source fragment appended and parsing is continued.
