The only uses of bzero() in the maiko sources come about because it is used in the macOS
definition of the FD_ZERO() macro in sys/select.h. Suppress the warning by including
-clang-analyzer-security.insecureAPI.bzero
in the clang-tidy list of checks. This may well have been replaced by a memset()
call in more up-to-date macOS releases.
* Rewrite keyboard and async I/O handling to improve performance and reduce dependency on async I/O signals
Replaces the SIGIO handler, which used to process X events in the interrupt context, with
a stub routine that sets a flag to signal processing is required.
Actual event processing is moved to the main dispatch loop where Lisp periodic interrupts
are handled.
Removes the X connection file descriptor from the set of fds contributing to SIGIO events
and moves the processing of X events to where the Lisp periodic interrupt is handled in
the main dispatch loop. This code is already guarded by a check for XPending() so can
be called regardless of whether any file descriptors are known to be ready.
Actual processing of async I/O events and X events are handled by procedures
process_io_events() and process_Xevents() respectively. For the most part these are
a renaming of getsignaldata() and getXsignaldata().
The Lisp periodic timer (VTALRM) was set to operate with a 25000 us period (40 Hz),
but on modern hardware it is possible to run this timer with a period of 10000 us (100 Hz)
Incidentally, a bug was noted (and fixed) in the X event handling code for motion events:
Mouse motion without any keyboard activity should not add an entry to the keyboard event
ring buffer as these events do not represent a key state change.
Since the ring buffer is of limited size, when it is filled new events are ignored
until the buffered events are processed. This resulted in the loss of a key/mouse button
transition (up or down) if the mouse was moved about "too much" between keyboard events.
A few incidental cleanups were also made:
- KBDEventFlg initialization fixed (wrong semantic type)
- Event_Req renamed to IO_Signalled (more appropriate name)
- int_io_open() sets up process (self) to handle SIGIO generated by O_ASYNC operations
- LOCK_X_EVENTS turned off since X library calls can no longer happen in an interrupt context
* Use of O_ASYNC must depend on the symbol being defined (looking at you, Cygwin)
* Add SA_RESTART flag to sigaction for SIGVTALRM periodic interrupt
* LOCK_X_UPDATE is no longer needed and should not be defined by default for Solaris in version.h
* Lock X updates for all systems and lock around X cleanup at exit
All systems which could generate I/O interrupts on the X file descriptor need
to have locking enabled, or there is a high probability that the X library
will deadlock. This requirement may be removed in the future with restructuring
of the way that the interrupt driven I/O is handled.
Ensure that the appropriate XLOCK/XUNLOCK is done around the X window
closing code.
* Also lock X updates when building with CMake.
* Remove code for unsupported Sun display configurations
Initial cleanup removing all code that was #ifdef'd for SUNDISPLAY.
Other SunWindows dependent code may also be removeable
* prropstyle will never be defined with no SunWindows support and therefore no pixrect code
* Remove code that is ifndef NOPIXRECT, since pixrects are part of the obsolete Sun windows code.
There are still traces of pixrect dependent code that could be removed.
On pushes to master: builds images for x86_64, Apple M1, and ARMv7
architectures based on Ubuntu Focal release. This builds only the
regular lde/ldex and does NOT build the ldeinit, so these are not
suitable for doing a loadup from scratch.
Commit based on https://github.com/stumbo/maiko/tree/gitHubBuild
getXsignaldata() can call additional Xlib procedures which could generate an I/O
interrupt. To avoid deadlocks in Xlib, ensure that XLocked is set when
getXsignaldata() is called.
* Remove unnecessary calls to XLOCK/XUNLOCK
The X methods called by flush_display_region() and flush_display_lineregion() handle the
locking/unlocking, therefore it is unnecessary to invoke the locking here.
* Remove explict signals when doing XUNLOCK()
Call getXsignaldata() directly if a signal happened while X code
was locked, instead of generating a signal and then handling it.
Some macros and functions needed to pass the DspInterface instead of
extracting the X display and window and passing those so that the
correct structure was available for the XUNLOCK() call.
The OPFN and OPFNX macros had been coded with a hand expansion of a loop
pushing values on the stack to separately handle the cases of 1, 2, and
more values. This actually lead to larger, likely slower, and certainly
less obvious code.
Until now, the only differentiation between using DLPI and using NIT for
the ethernet interface was expressed as a function of USE_DLPI.
This commit makes explicit when code is for the DLPI interface or the NIT
interface, with USE_DLPI and USE_NIT. This is setup for using the BPF
interface to the ethernet as an alternative.
* Fix buffer overrun vulnerability: use strncpy
read_Xoption uses a char buffer defined in main.c with length
MAXPATHLEN, aka PATH_MAX in POSIX. Unfortunately it was using strcpy to
copy from the command-line arguments (via argv) and the environment (via
getenv) without any bounds checking whatsoever. This could very easily
cause an overflow.
It's unlikely that a user will want to provide a path longer than
PATH_MAX-1 (a generous 1023 bytes on my machine). If they try, we should
stop them from causing any damage.
* Use strlcpy instead of strncpy
Thanks to Nick Briggs for the suggestion. It would be best to use
sizeof(sysout_name) instead of hardcoding a reference to the PATH_MAX
constant, but unfortunately sysout_name is an extern in xrdopt.c and so
the compiler doesn't know its size. I don't want to mess with that
coupling in this commit, because I assume there was a reason for doing
it that way rather than putting sysout_name in a header; I'll keep the
scope of the changes here small.
* Revert "Use strlcpy instead of strncpy"
Ah. This is not great. Turns out strlcpy is a nonstandard BSD extension
with its own set of problems
[https://en.wikibooks.org/wiki/C_Programming/C_Reference/nonstandard/strlcpy]
that we may be best served by avoiding. On Linux, it's only accessible
through libbsd, and we have no other reason (as far as I can tell) to
require that. Unless we want to provide our own strlcpy implementation,
we should stick with strncpy. It's far safer than what was there before
and doesn't present any edge cases in this scenario that are apparent to
me.
* Display atom name in error message if get_package_atom() fails
* Various fixes to package/atom handling in testtool.c
Remove S_TOPVAL and S_MAKEATOM which only existed to deal with an old issue
with dbx where you supposedly couldn't enter a string with "\" in it.
Remove countchar(), which is functionally identical to strlen(), and adjust
code that used it.
Adjust return type of MAKEATOM() to be the LispPTR that it should be, instead of int.
Limit find_package_from_name() to examining only the number of entries that are
present in the *PACKAGE-FROM-INDEX* array, instead of walking off the end.
MakeAtom68k() now drops into uraid() if asked to look up an atom that does not exist
(Make... is a misnomer, it will never *make* the atom, only lookup an existing
one)
Except where the expansion would be syntactically invalid,
for example "goto macroarg;" detection of which is a bug in clang-tidy,
so warn it off with a NOLINT...(bugprone-macro-parentheses)
* Remove unused `print_lispusage()`.
* Make some globals into function locals.
* Make the options table into a `static`. It can't be `const`
because `XrmParseCommand` wants a mutable pointer.
* Remove unused `homeDB`.
* Stop using `caddr_t`, use `XPointer`. `caddr_t` is not POSIX
and never made it past old obsolete BSD code.
* Remove unused `Master` argument.
* Update documentation.
* Remove unused `MasterFD` variable in caller.
* Remove `slot` variable in caller and use `Master` in each case.
* Make `FindAvailablePty` a static function.
