This resolves all issues the test suite uncovered. It should also deal
with hard links correctly, though that hasn't been tested. Still
unsupported are:
* changes due to mounting/unmounting a volume,
* tracking of symlinks in the path components.
Add inner class BWatchingInterface and method SetWatchingInterface().
This abstracts the calls to watch_node() and stop_watching(), thus
making it possible to use the path monitor in Tracker.
The main point of this commit is to fix this line:
temp << (char)type << "' got '" << token.string << "'";
which gets printed when DeskCalc encounters a parse error.
Specifically the (char)type part needed fixing.
This code would try to print the char equivalent of a token which got
converted to lower ascii character between 0 and 15. This would at
best result in a newline and never anything helpful.
I took the germ of idea and expanded upon it reassigning the TOKENs to the
numeric values of the printable characters they represent where applicable.
For instance TOKEN_STAR now has a value of 42 which is ascii for '*'. By
using implicit char -> int conversion the numeric value is avoided in the code.
So now (char)type will, in many cases get you the equivalent ascii char
represented by that type. Those that don't such as TOKEN_IDENTIFIER and
TOKEN_CONSTANT are special cased.
Once the TOKEN's values correspond to their ASCII equivalents some other
simplifications became possible interchanging the TOKEN and the character
it represents.
This targets a problem where a numbers with large numbers of non-decimal
significant digits took a long time to round after converting to scientific
notation because they are rounded one character a time.
To solve this, after converting to scientific notation lop off everything
after 40 characters greatly reducing the amount of further rounding
needed.
An example I used to test this was to calculate 10,000! which gives
a result with 35660 significant non-decimal digits (aka a lot). By loping
off numbers after 40 characters before rounding to fit the operation goes
from ~10 seconds to complete to under a second.
I chose 40 as a max as it is large enough to ensure that the result will
get rounded with some leeway provided for font width variations.
Worse-case scenario is the result is off by 1 in the last place.
Numbers with large numbers of significant decimal digits get rounded
by MAPM so aren't a problem.
Don't release the reference to the current block until we get the
notification that the next block has been retrieved. Otherwise, the
previous/next block buttons would cease to work if the requested block
failed to be retrieved.
- In the case where retrieval of a memory block failed, InspectorWindow
didn't handle the notification. Consequently, it never removed itself as
a listener from the failed block, nor did it release its reference for
it. Consequently, if one attempted to retrieve data from the same block
again, walking the listener list would crash due to the already-deleted
entry in the list.
- The success case had the same problem with regards to not removing its
listener, but was masked by virtue of the inspector currently being the
only user of the memory block manager, so in the latter case the blocks
would be properly released/destroyed and the aforementioned walk would
never occur.
- Adjust locking a bit to ensure that manipulating the listener list
always happens with the team lock held.
- Style fixes.
We can't use BPath to get the filename for the source file path embedded
in the debug information, since it may be relative, which BPath will try to
normalize.
* added a copyright header.
* fixed includes order.
* don't return early on B_BAD_SEM_ID as it means the writer has finished,
but there can be data to read.
* free resources in acpi_namespace_free() instead of acpi_namespace_close().
This reverts commit ea27e95f489fbb29cedad74788ee607b331f8a2f.
* AnEvilYak pointed out that this was a false positive as
BObjectList can optionally delete on remove.
* I'll add a penny to the bitcoin bad commit jar :)
* The config space is larger than 255, we need to use an uint16 to access
offsets superior or equal to 256. The current API only proposes an uint8 for this.
This change switches the offset parameter to the uint16 type. Axel hinted that
the used values are the same with such a change (the doc says sign extended to 2 or
4 bytes).
I checked with GCC2 and it's indeed the case when inspecting the memory.
With GCC4, instructions are the same on function call.
* prints info about extended capabilities.
* struct pci_module_info and struct pci_device_module_info are extended with
pci_find_extended_capability().
* In case the locale backend could not be loaded, these functions (and
their reentrant counterparts) just returned an error. So we reactivate
parts of the BSD-/Olson-implementation in localtime_fading_out.c in
order to use them as fallback.
* Cleanup localtime_fading_out.c (remove a lot of unused cruft).
* all those functions need to return the given wc unchanged in case of
error, not 0
* towctrans() didn't actually look at the requested transition, but
always acted as if _ISlower was given
- When possible, SourceView now adds a context menu option to switch
between source and disassembly. If the disassembled code is not yet
available, it is asynchronously requested. Adjusted SourceView::Listener
and implementing subclasses accordingly to make that request feasible.
- Adjust TeamWindow to correctly deal with the possibility of the
function source code being available but not loaded.
