* Force building Lua rocks instead of trying to install a binary.
* Set the CC and LD environment variables to pass
CMAKE_C_COMPILER_ARG1 (contains `-m32` for 32-bit builds)
to `luarocks build`.
It turns out that Busted started cleaning the environment in 2.0rc5 as a
result of Olivine-Labs/busted#62. This, in turn, caused the ffi module
to be reloaded for each spec file, and LuaJIT doesn't appreciate it.
The net effect is an assertion error in LuaJIT.
By using the --helper feature of Busted, we can pre-load some modules
ahead of Busted and prevent it from reloading them--making LuaJIT happy
again.
This is necessary for newer versions of Busted, otherwise assert will be
nil and the tests will die.
Note: this does not mean the tests now work with the latest Busted.
There are still several issues preventing that from happening.
busted depends on a number of libraries that sometimes introduce crashing bugs.
The new custom target makes sure only stable versions of those dependencies are
installed.
Problem : Branch condition evaluates to a garbage value @ 2868.
Diagnostic : False positive.
Rationale : Array has_hotkey, declared at 2812, is initialized by
console_dialog_alloc (only the needed number of elements).
That same number of elements is used by
copy_hotkeys_and_msg.
Suggested path error is impossible, because it involves a
different number of elements in those functions.
Resolution : Above condition is cumbersome to prove through assertions.
Thus, we prefer to just initialize the array to all-false
at declaration point before calling console_dialog_alloc.
- process spawning was decoupled from the rest of the job control logic. The
goal is reusing it for spawning processes connected to pseudo terminal file
descriptors.
- job_start now receives a JobOptions structure containing all the startup
options.
Problem : Use after free @ 1795.
Diagnostic : Real issue.
Rationale : prev_curtab can in fact be freed as a result of call
`win_close_othertab(win, free_buf, prev_curtab);`, but it's
later used at
`sprintf(..., tabpage_index(prev_curtab));`.
This was introduced at
3ffc5d81c3.
Resolution : Move prev_idx calculation before the call freeing
prev_curtab.
Problems : Assigned value is garbage or undefined @ 5363.
Result of operation is garbage or undefined @ 5356.
Result of operation is garbage or undefined @ 5320.
Result of operation is garbage or undefined @ 5192.
Diagnostic : False positives / Multithreading issues.
Rationale : Suggested error paths contain incoherent values for
has_mbyte, enc_utf8, and enc_dbcs, which should always hold
the relationship has_mbyte = enc_utf8 || enc_dbcs, with
enc_utf8 and enc_dbcs being mutually exclusive.
Asserting on the globals, though, fails, because checker
believes they could be modified by other threads in
between.
Resolution : Make local copy of globals and assert relationship on them.
The current will segfault for large chunks of output because the output buffer
will be overrun.
Using unibi_format is simple because we can simply flush the buffer when its
full.
Many common terminals that set TERM=xterm and $COLORTERM support 256 colors. If
this is detected, use force the hardcoded xterm's setaf/setab capabilities.
- Removed term.c, term.h and term_defs.h
- Tests for T_* values were removed. screen.c was simplified as a
consequence(the best strategy for drawing is implemented in the UI layer)
- Redraw functions now call ui.c functions directly. Updates are flushed with
`ui_flush()`
- Removed all termcap options(they now return empty strings for compatibility)
- &term/&ttybuiltin options return a constant value(nvim)
- &t_Co is still available, but it mirrors t_colors directly
- Remove cursor tracking from screen.c and the `screen_start` function. Now the
UI is expected to maintain cursor state across any call, and reset it when
resized.
- Remove unused code
Problem : Out-of-bounds access @ 3730.
Diagnostic : Real issue.
Rationale : str is constructed step by step, str_l growing each time.
str_m is the maximum length of str. So, at every step,
avail is computed to see if the piece to be added fits in.
If not, piece is truncated to a max of `avail`, so that str
stays in bounds. Such blocks where pieces are added are of
the form `if (str_l < str_m)`. It then follows that once
one of those pieces exhausts available space on str, no
other such block should be entered. Formally:
str_l < strl_m && avail = str_m - str_l && x >= avail
-->
str_l + x >= str_m
Now, suggested error path successively enters blocks where
str is exhausted. We're not sure if coverity just fails to
follow above implications, or, on the contrary, it's aware
of them, but it's signaling the more complex possibility of
implications not being fulfilled because of possible
arithmetic overflows. We opt then to assume this last case,
as the possibility is in fact there.
Resolution : Refactor code so that tracked condition doesn't depend on
arithmetic implications. Check for overflow.
