We currently default to "-mfloat-abi=hard -mfpu=fpv4-sp-d16" for M4F cores, and
and variations of "-mfloat-abi=soft" for the others. Keep the M4F default, and
move others to no FP flags for consistency, but allow overriding these flags
via the FP_FLAGS environment variable.
Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
After a STALL handshake is transmitted, a control pipe becomes idle. Not
marking the pipe as idle did not affect the STM32 family. Since it
distinguishes between OUT and SETUP tokens, it calls the setup handler
on a SETUP token, regardless of the state of the pipe.
Other families, such as LM4F do not distinguish in software between IN and
SETUP tokens, and need to decide which handler to call based on the state
of the pipe. On these chips, SETUP transactions will not be handled
properly after a transfer was STALLED, as the state machine of the pipe is
b0rked. Unb0rk it.
Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Each thread runs its own stylecheck script, outputing to {FILENAME}.stylecheck file. If that file is empty, it is deleted. Otherwise it is printed to the output and stays in the sources directory to indicate something in that file is not correct.
The file mangling and cat-ing is needed because of multithreaded nature of paralell make. (We like single file to be an block of text on the output).
New PHONY added that make styleclean clears all .stylecheck files.
New functionality added to check only single file (specifyilng the file with stylecheck extension) ie
make lib/usb/usb.c.stylecheck
BUG: if someone edit that file, .stylecheck file is not regenerated - missing dependency on the %.stylecheck rule. FIX: make styleclean stylecheck rebuild all style errors
Performance:
make stylecheck 1:43 -> 1:54 (slightly slower due to file creation)
make stylecheck -j8 1:43 -> 0:34 (Rapidly faster)
note the -j option must be specified with number, OS Windows cannot handle one thread per file.
make clean 34.64s -> 35.35s
make clean -j 14.00s -> 14.16s
(measured on Core2 quad, 2.84GHz, 4GB ram, windows x86)
Added overhead is logging the operation with every yamlfile to output.
Because most changes are internal to a function in exti.c I am leaving
the macro checks in there, otherwise we would end up with a bunch of
code duplication if we used the dispatch system used otherwise
throughout the stm32 part of the library. I bet it could be split up
into more granular functions resulting in more generic code that we
could run through the dispatch system. But I am leaving that as an
excersize for later.
We don't support f0 yet so let's not fool anyone. We may rename those
files back again if when we cross check that it is actually true this
file supports f0.
BuFran
Ben Gamari
Piotr Esden-Tempski
Alexandru Gagniuc
Dirk Braun