could prove to be a major PITA (because of all the dependencies that would need to be rebuilt)
* when adding a library to the image, its optional minor abi version is taken care of, too (i. e. a
corresponding link is created).
* the ABI-related links in /system/lib are now replicated in /boot/develop/lib/<arch>
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@35362 a95241bf-73f2-0310-859d-f6bbb57e9c96
build system
* declared ICU-libs as private system libraries, which causes them to no longer
be available for building software (they no longer are linked from the
development lib folder, so the linker won't find them)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@35323 a95241bf-73f2-0310-859d-f6bbb57e9c96
tree within qt-creator (at least auto-completion and navigation seem to work
fine)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@35074 a95241bf-73f2-0310-859d-f6bbb57e9c96
default address of the outgoing interface as a side-effect
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34989 a95241bf-73f2-0310-859d-f6bbb57e9c96
exposes a problem in our UDP-connect implementation: the local address is
not set as a side-effect of connecting to a specific destination.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34987 a95241bf-73f2-0310-859d-f6bbb57e9c96
would yield double inversion of the caret position, causing #5006
* cleanup with respect to when the selection can be highlighted and/or the
caret can be shown/hidden
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34944 a95241bf-73f2-0310-859d-f6bbb57e9c96
(in libgnu.so) with my current rsync port:
* Node now opens its fd with O_RDONLY, as otherwise BFS will refuse to open
(the attributes of) directories
* Node::Get() and Node::Set() now make use of the encoded attribute type,
as otherwise all created attributes would have the type 'XATR' (instead
of the encoded haiku-specific type)
* minor cleanup with respect to the line width limit
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34904 a95241bf-73f2-0310-859d-f6bbb57e9c96
'+' as part of their name, which is illegal for shell variables (where the
corresponding replacement character is 'X').
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34903 a95241bf-73f2-0310-859d-f6bbb57e9c96
(--update must be given as first option, but can be followed by others)
* added simple perl scripts that is used by configure to convert BuildSetup
from jam to shell format
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34896 a95241bf-73f2-0310-859d-f6bbb57e9c96
* let BuildSetup set HAIKU_HOST_IS_BIG_ENDIAN according to HOST_GCC_MACHINE
* minor cleanup in BuildSetup
* when building ICU, set the endianness indicator suffix according to
HAIKU_HOST_IS_BIG_ENDIAN
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@34864 a95241bf-73f2-0310-859d-f6bbb57e9c96
* make word-wise navigation via cursor keys match Pe, such that it is consistent
throughout haiku (unless some applications brew their own)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33913 a95241bf-73f2-0310-859d-f6bbb57e9c96
which are now being used to implement the word-wise keyboard navigation,
fixing #4785
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33885 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Word selection via double-click now works like in R5, except that '_' is now
being considered part of a textual word, too. This fixes#4754.
* Adjusted CanEndLine() to make much more sense (at least for western
languages).
While I have tested the wrapping quite a bit, it is not unlikely that these
changes introduce regressions. If you spot one, please tell!
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33882 a95241bf-73f2-0310-859d-f6bbb57e9c96
fixes#4894
* PointAt() now rounds x- and y-positions to the next integer instead of always
rounding them up - this yields improved caret-positions (e.g. before an 'f'
character)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33877 a95241bf-73f2-0310-859d-f6bbb57e9c96
* copied implementations for llround(), llroundf() and llroundl() from
glibc-2.3.2
* added corresponding declarations to math.h
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33863 a95241bf-73f2-0310-859d-f6bbb57e9c96
(reported by Axel):
* _AutoResize() can no longer check the difference between the textrect and
the maximum line width, as the textrect is no being adjusted earlier during
the line-calculation code - instead, it needs to measure the difference
between the bounds and the textrect width plus the insets.
* MakeResizable(true) needs to reset the right inset, as otherwise the auto-
resizing would get confused about how large the textview needs to be
(seems to be an artefact of how Tracker creates the textview).
* GetInsets() doesn't need to compute the insets, as they're maintained in
fLayoutData.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33813 a95241bf-73f2-0310-859d-f6bbb57e9c96
* SetTextRect for non-wrapped textviews needs to update the insets,
as otherwise several misalignments occur within several apps - renaming
in Tracker was effected the worst, though
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33717 a95241bf-73f2-0310-859d-f6bbb57e9c96
during the width calculation
* used _TabExpandedStyledWidth() at all places that need to expand tabs,
fixing #4822
* removed _StyledWidthUTF8Safe() as it didn't make sense - we now perform
the same auto-scrolling horizontally as we do vertically
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33713 a95241bf-73f2-0310-859d-f6bbb57e9c96
* switching from wrapping to non-wrapping mode triggers a recomputation of
the text rect
* non-wrapping mode now triggers updating of the data rect for the horizontal
scrollbar as needed, fixing #4705
* corrected a couple of width computations that confused bounds and text
rect width, leading to drawing artefacts at the end of a line
* unified computation of text rect width to always take all lines into
account (not just the first one)
* some cleanup
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33703 a95241bf-73f2-0310-859d-f6bbb57e9c96
to avoid recurring problems during migration of subversion checkouts
(restored binary files that were garbled by subversions during checkout)
* added appropriate svn:mime-type property for problematic (binary) files
* removed a single (mistyped) svn:mimetype property
* dropped svn:eol-style property for cleanup (they all contained 'native')
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33670 a95241bf-73f2-0310-859d-f6bbb57e9c96
software looking for them will find them) as a temporary fix for #4747.
In the long run, the functions declared in these headers should be
implemented by means of ICU and then the headers shall be made public again.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33575 a95241bf-73f2-0310-859d-f6bbb57e9c96
It is good practive to #undef this before every definition,
due to how the ExtractCatalogEntries jam-rule works.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33360 a95241bf-73f2-0310-859d-f6bbb57e9c96
* fixed typo and "correction" in de.catkeys (it is not
possible to change the english base text this way - it
will corrupt the catalog instead, as the fingerprint
will no longer match)
* fixed typo in LocaleWindow.cpp and adjusted the existing
catkeys accordingly
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33350 a95241bf-73f2-0310-859d-f6bbb57e9c96
work to do, but it's about time to give this code more exposure.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33176 a95241bf-73f2-0310-859d-f6bbb57e9c96
Net_server starts services by invoking fork() followed by exec(). If the latter
fails (for instance because the service isn't installed), the forked child is
invoking exit(). This in turn unloads libbe, triggering static cleanup code in
BMessage, which deletes a couple of message ports that were inherited from the
parent during the fork. After that, net_server was desparately missing those
ports and no longer worked reliably.
* in InitTerminateLibBe, we now register an atfork-(child-)handler, which
takes care to re-initialize the static reply ports used by BMessage code
* added BMessage::Private::StaticReInitForkedChild wrapper and
BMessage::_StaticReInitForkedChild() implementation which overwrites the
inherited port IDs with a set of own ports
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33050 a95241bf-73f2-0310-859d-f6bbb57e9c96
as a result failed DHPC-requests will no longer result in bogus static
configurations (but an auto-config IP-address will be picked instead)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33049 a95241bf-73f2-0310-859d-f6bbb57e9c96
its open_dir() implementation instead (as suggested by Ingo).
-alphabranch (it's only a cleanup)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33025 a95241bf-73f2-0310-859d-f6bbb57e9c96
* update Firefox optional package for r1a1
+alphabranch (will do that next)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33019 a95241bf-73f2-0310-859d-f6bbb57e9c96
* fixed precedence problem in jam rule that caused an if to evaluate to true, no matter
if a cd or an image was being built
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@33007 a95241bf-73f2-0310-859d-f6bbb57e9c96
a directory, we returned EINVAL instead, which caused Perl's internal
glob() implementation to fail prematurely
+alphabranch (will do that myself in a minute)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@32999 a95241bf-73f2-0310-859d-f6bbb57e9c96
longer require libiconv
* removed libiconv from dependencies of DevelopmentMin, as this is
no longer true
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@32479 a95241bf-73f2-0310-859d-f6bbb57e9c96
although it should have been safe before, too, since BTextView neither
has inline methods nor any friends that live outside of libbe
* added new member fLastClickOffset in order to be able to restore the
triple-click handling (linewise selection)
* renamed fClickOffset to fCaretOffset, as that is what it really is
now
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@32444 a95241bf-73f2-0310-859d-f6bbb57e9c96
* when items are removed, the anchor may need adjustment (fixes the problem
in Beam reported by Stippi)
* the check for needed adjustment of fLastSelected in _Select(from, to, extend)
was broken, as it needs to be independent of fFirstSelected - this fixes
broken navigation after doing a select-all
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@32099 a95241bf-73f2-0310-859d-f6bbb57e9c96
* avoid setting fAnchorIndex multiple times in certain situations
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@32090 a95241bf-73f2-0310-859d-f6bbb57e9c96
* fixed handling of B_HOME and B_END in extend mode for multiselection, such that
now all the items inbetween the anchor and the final target are selected, too
* made sure that the anchor always gets adjusted in _Select(), as otherwise
navigation via cursor keys could get "locked".
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31980 a95241bf-73f2-0310-859d-f6bbb57e9c96
* fixed recent regression with respect to navigating via cursor keys
* unified (and thus partly corrected) computation of whether or not we're in
extend mode in KeyDown(), fixing the handling of unmodified B_HOME and B_END
in multiselection mode
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31975 a95241bf-73f2-0310-859d-f6bbb57e9c96
* resolved a couple of redundant function declaration in our public headers
* adjusted zipomatic accordingly - everything else built fine
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31938 a95241bf-73f2-0310-859d-f6bbb57e9c96
those, too. Among other things, this should facilitate making use of wstring
with gcc4.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31927 a95241bf-73f2-0310-859d-f6bbb57e9c96
* added declaration of strtof() to stdlib.h, as we now provide it alongside
with strtod()
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31926 a95241bf-73f2-0310-859d-f6bbb57e9c96
too, and I do not dare to add a glibc-based source to our kernel.
This means that our kernel will have to live with strod() yielding
inaccurate results, but since the only use I have found was for fs-queries,
I think we can cope for now.
* added a comment to the top of strtod.c that explains the current state.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31925 a95241bf-73f2-0310-859d-f6bbb57e9c96
version that lives in our repo, too, and actually works as intended
* removed the broken strtod() implementation from the repo
This fixes all floating-point test regressions reported in #3308.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31919 a95241bf-73f2-0310-859d-f6bbb57e9c96
all allocations, instead we now directly request the use of malloc/free in
stl_config.h. This avoids warnings about 'arrays of size 0' when building
any stl-code with -pedantic (like e.g. ICU does).
Alternatively, we could active pthread-type threading support, but since in
some small tests, I have not been able to see any advantage, I've just added
a respective comment.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31609 a95241bf-73f2-0310-859d-f6bbb57e9c96
in /boot/common/bin (why does it care where sed lives?)
* revert subversion to 1.6.2, as that does not depend on the expat optional
package
I think we (well, I) should be more thorough when testing new packages,
it would probably be a good idea to test each package on a new, minimal
image.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31486 a95241bf-73f2-0310-859d-f6bbb57e9c96
it's signature indicates that it should - the callers just never use
the value currently, so it caused no harm
* squashed a TODO in the runtime_loader about the resolving strategy
for undefined symbols in add-ons: I've implemented the breadth-first
strategy (leaving out the add-on itself), as that one made most sense to
me.
This avoids loader problems with older optional packages of Pe on gcc4, as some
add-ons (e.g. "Expand Tabs") could not be loaded due to them containing undefined
symbols that are found in second level dependencies (libstdc++.so in this case).
Ingo: please review.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31465 a95241bf-73f2-0310-859d-f6bbb57e9c96
* the default ABI for gcc2 builds should be B_HAIKU_ABI_GCC_2_HAIKU, not B_HAIKU_ABI_GCC_2, as the latter
is equal to B_HAIKU_ABI_GCC_2_ANCIENT and enforces the old symbol resolving behaviour.
If our build isn't ..._HAIKU, what is? ;-)
This fixes a whole lot of perl tests, which failed because the required shared libs could not be loaded
(as they contained undefined symbols that are only found in second level dependencies).
Note: the existing perl optional package works, since it does not contain any ABI specification symbol yet.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31464 a95241bf-73f2-0310-859d-f6bbb57e9c96
keeps its own idea about what a wchar_t is and that was still a short.
This of course made the data structure of a streambuf incompatible with the
"lower" half - the glibc part - causing (potentially all sorts of) crashes
when these classes were used.
This should fix the crash of gensyscalls when building haiku natively
on a gcc2-haiku.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31462 a95241bf-73f2-0310-859d-f6bbb57e9c96
it (headers and library) from the image. The libiconv in our tree is
only used internally as a backend for libtextencoding. The real libiconv
is provided as an optional package.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31452 a95241bf-73f2-0310-859d-f6bbb57e9c96
* gcc-4.3.3 did contain too-far-stripped versions of static libraries, missing all symbols.
Now only the unneeded ones have been removed - fixing the build of the kernel on haiku natively
* libiconv was missing the shared versions of the libraries. I have updated the package to 1.13.1
and built it with shared libs (which took me considerable time to get right - the auto*-suite
basically sucks)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31450 a95241bf-73f2-0310-859d-f6bbb57e9c96
happens after the tokens have been declared - this fixes#3137 for good
(without ftpcmd.y depending on itself)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31401 a95241bf-73f2-0310-859d-f6bbb57e9c96
of trunk]
Fixed a bug exposed by changes in the locale kit branch:
* the signature of stringbuf::setbuf() wasn't matching the one from streambuf,
which caused the base version to be hidden instead of overridden - leading
to crashes during cleanup of stringstream objects
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31262 a95241bf-73f2-0310-859d-f6bbb57e9c96
* added WCHAR_MIN and WCHAR_MAX - already as 32-bit values, since that's
what wchar_t is going to become, soon (and those values could not have
been used before, since they were missing)
* this is still work-in-progress though, I need to run more tests to find
out which of these functions are actually working, too
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31182 a95241bf-73f2-0310-859d-f6bbb57e9c96
several problems in our wchar-support - the worst of which being that
our compilers define wchar_t as short while the glibc-implementation
is relying on it being 32-bits wide ...
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31180 a95241bf-73f2-0310-859d-f6bbb57e9c96
next position of the caret and only scroll the distance between the current
and the next position. This fixes#3981
* When paging upwards, we need to compensate for the fact that the caret
position is always considered at the top of the line, as otherwise a
page-up would pass one more line than a page-down
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30933 a95241bf-73f2-0310-859d-f6bbb57e9c96
* when selecting via mouse, the caret position now follows the direction of
the selection - which is less surprising than not moving the caret at all
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30928 a95241bf-73f2-0310-859d-f6bbb57e9c96
haiku would support any of the declared functions (none of which are
provided by libroot)
These will come back once the locale kit has been integrated
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30883 a95241bf-73f2-0310-859d-f6bbb57e9c96
on an address in host-byte-order, not network-byte-order.
Basically this caused failure for all IP-traffic to any IP-address
where the last byte & 0xF0 == 0xe0.
Now we can surf www.pcre.org again ;-)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30870 a95241bf-73f2-0310-859d-f6bbb57e9c96
(3-letter) and numerical timezones (e.g. -0700) are supported
This means that the workaround in mail_util.cpp of the mail kit could be
removed, but since I do not know if that should still support other versions
of BeOS, I have chosen to leave it alone.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30864 a95241bf-73f2-0310-859d-f6bbb57e9c96
* applied a patch suggested by kaliber that seems to fix the issue for good,
thanks!
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30861 a95241bf-73f2-0310-859d-f6bbb57e9c96
* added TODO about a sanity check that should perhaps be dropped
I checked this change with both a haiku-BFS partition and one from R5
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30852 a95241bf-73f2-0310-859d-f6bbb57e9c96
and the signal stack must be copied into the new thread (at least that's
what I could make of IEEE Std 1003.1-2001 regarding the expected behaviour
of fork)
This brings down the failures of the rsync test suite from 24 to 18 ...
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30807 a95241bf-73f2-0310-859d-f6bbb57e9c96
Ingo / Axel: please shout if the Advantage isn't correct - I could only test
with my Ergo Elan (the international version).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30803 a95241bf-73f2-0310-859d-f6bbb57e9c96
made non-selectable on program start and selectable after the credits have
been added escapes me.
This fixes#2497 again.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30796 a95241bf-73f2-0310-859d-f6bbb57e9c96
Oliver Tappe