| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
|/
|
|
| |
Suggested by svaarala.
|
|
|
|
|
|
|
| |
Some features of later versions are used but are either restricted to OS4 only, or runtime version checks select the appropriate routines.
Where the lowest required version cannot be determined, the version provided in the ClassAct 3.3 archive is requested.
Bugs may mean these older classes do not work as expected.
In addition, restrict the opening of listbrowser.gadget and the custom Stringview class to OS4 only.
|
| |
|
|
|
|
| |
Restore 6d63f7959af64a45b0643d0610fcbdb0c07bfbc4 for 2.1.0 import.
|
|
|
|
| |
Restore 336326af3aab93f31474fa6de28782457ae4a1c0 for 2.1.0 import.
|
| |
|
|
|
|
|
| |
When we aren't modifying the line, pango_layout_get_line_readonly()
is a faster alternative to pango_layout_get_line().
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
With this change we are consistent about how the pango layout
we use is created.
Now it always comes from a pango_layout_new() call on a pango context
that comes from gdk_pango_context_get().
Previously the pango layout used for painting came from a call
to pango_cairo_create_layout(), which required a global called
"current_cr" (a cairo drawing context), which is only valid
during redraw (painting). Since it was only valid during
painting, this source could not be used for pango layout
creation for the measuring code.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Although the length is unused when margin is auto, having
uninitialised values present in a computed style means that
the style hashing may give the same style different hashes
depending on the unintitialised info.
The effect of this would have been be to reduce the chance
of computed style sharing. It would have had no effect on
page rendering.
|
|
|
|
|
|
|
|
|
|
|
|
| |
When assert() is called, which is not uncommon in utility code within NetSurf,
we lose coverage data for anything done before the assert() in the test. This
commit corrects that oversight but is at least slightly GCC specific and may
need tweaks for non-Linux platforms.
By default, 'make coverage' will enable assert coverage, and it can be disabled
with 'make coverage NOASSERTCOVERAGE=yes' if necessary.
Signed-off-by: Daniel Silverstone <dsilvers@netsurf-browser.org>
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
build/Linux-atari/content_handlers_javascript_duktape_duktape.o:build/Linux-atari/content_handlers_javascript_duktape_duktape.o:(.text+0x978): undefined reference to `_trunc'
build/Linux-atari/content_handlers_javascript_duktape_duktape.o:build/Linux-atari/content_handlers_javascript_duktape_duktape.o:(.text+0x990): undefined reference to `_log2'
build/Linux-atari/content_handlers_javascript_duktape_duktape.o:build/Linux-atari/content_handlers_javascript_duktape_duktape.o:(.text+0x99c): undefined reference to `_cbrt'
|
|
|
|
| |
Restore 6d63f7959af64a45b0643d0610fcbdb0c07bfbc4 for 2.0.2 import.
|
|\ |
|
| |
| |
| |
| | |
Restore 336326af3aab93f31474fa6de28782457ae4a1c0 for 2.0.2 import.
|
| |
| |
| |
| |
| |
| | |
https://github.com/svaarala/duktape/blob/master/doc/release-notes-v2-0.rst#duk_error-duk_error_va-duk_throw-duk_fatal-have-a-return-value
This also changes Console.bnd to return on error, rather than ignoring it.
|
| |
| |
| |
| | |
https://github.com/svaarala/duktape/blob/master/doc/release-notes-v2-0.rst#duk_safe_call-userdata
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Note that DUK_OPT_XXX feature selection has been removed.
This means we can't use `DUK_OPT_HAVE_CUSTOM_H`, so we need another
way to tell duktape about our custom header.
This means assembling our own duktape distribution with a
duktape "tools/configure.py" invocation.
Instructions for updating duktape have been added to NetSurf's
`Docs/` directory.
See:
https://github.com/svaarala/duktape/blob/master/doc/release-notes-v2-0.rst#duk_opt_xxx-feature-option-support-removed
|
|
|
|
|
|
|
| |
Previously, aborted HTML fetches left the content in LOADING state,
and the hlcache clean code does not clean contents in the LOADING
state. They, and their child contents ended up stuck in the
cache forever.
|
| |
|
|
|
|
| |
This makes it easier to track a content by its ID through a log.
|
|
|
|
|
|
|
|
|
|
| |
This builds a dynamic library which can cause heap allocation (malloc)
faliures after a specified number of calls. This is useful to allow
tesing of memory failure allocation paths within netsurf test suites.
An example test for core strings test has been added which
incrementaly fails alloctions allowing all error paths in
initialisation to be exercised.
|
| |
|
| |
|
| |
|
| |
|