1 #This configuration file is similar to a regular .ini file. Comments start
2 #with hashtags or semicolons, sections are written in square brackets and
3 #in each section there can be arbitrary many key-value pairs.
5 #There are 3 sections currently: ‘flags’, ‘warnings’, ‘optimizations’.
6 #They contain a list of boolean values of the form ‘VARNAME = true’ or
7 #‘VARNAME = false’. The variable names are the same as for the corre‐
8 #sponding -W, -f or -O flag written with only capital letters and dashes
9 #replaced by underscores.
12 #Add some additional characters to the string table in order to
13 #compensate for a wrong boundcheck in some specific version of the
16 DARKPLACES_STRING_TABLE_BUG = true
19 #When assigning to field pointers of type .vector the common be‐
20 #haviour in compilers like fteqcc is to only assign the x-compo‐
21 #nent of the pointer. This means that you can use the vector as
22 #such, but you cannot use its y and z components directly. This
23 #flag fixes this behaviour. Before using it make sure your code
24 #does not depend on the buggy behaviour.
26 ADJUST_VECTOR_FIELDS = true
29 #Enable a partially fteqcc-compatible preprocessor. It supports
30 #all the features used in the Xonotic codebase. If you need more,
36 #Enable some predefined macros. This only works in combination
37 #with '-fftepp' and is currently not included by '-std=fteqcc'.
38 #The following macros will be added:
50 #Note that __FUNC__ is not actually a preprocessor macro, but is
51 #recognized by the parser even with the preprocessor disabled.
53 #Note that fteqcc also defines __NULL__ which becomes the first
54 #global. Assigning it to a vector does not yield the same result
55 #as in gmqcc where __NULL__ is defined to nil (See -funtyped-nil
56 #), which will cause the vector to be zero in all components. With
57 #fteqcc only the first component will be 0, while the other two
58 #will become the first to of the global return value. This behav‐
59 #ior is odd and relying on it should be discouraged, and thus is
60 #not supported by gmqcc.
65 #Enable math constant definitions. This only works in combination
66 #with '-fftepp' and is currently not included by '-std=fteqcc'.
67 #The following macros will be added:
84 FTEPP_MATHDEFS = false
87 #Allow switch cases to use non constant variables.
92 #Perform early out in logical AND and OR expressions. The final
93 #result will be either a 0 or a 1, see the next flag for more pos‐
99 #In many languages, logical expressions perform early out in a
100 #special way: If the left operand of an AND yeilds true, or the
101 #one of an OR yields false, the complete expression evaluates to
102 #the right side. Thus ‘true && 5’ evaluates to 5 rather than 1.
107 #Enable the underscore intrinsic: Using ‘_("A string constant")’
108 #will cause the string immediate to get a name with a "dotrans‐
109 #late_" prefix. The darkplaces engine recognizes these and trans‐
110 #lates them in a way similar to how gettext works.
112 TRANSLATABLE_STRINGS = true
115 #Don't implicitly convert initialized variables to constants. With
116 #this flag, the const keyword is required to make a constant.
118 INITIALIZED_NONCONSTANTS = false
121 #If this flag is not set, (and it is set by default in the qcc and
122 #fteqcc standards), assigning function pointers of mismatching
123 #signatures will result in an error rather than a warning.
125 ASSIGN_FUNCTION_TYPES = true
128 #Produce a linenumber file along with the output .dat file.
133 #Use C's operator precedence for ternary expressions. Unless your
134 #code depends on fteqcc-compatible behaviour, you'll want to use
137 CORRECT_TERNARY = true
140 #Normally vectors generate 4 defs, once for the vector, and once
141 #for its components with _x, _y, _z suffixes. This option prevents
142 #components from being listed.
145 SINGLE_VECTOR_DEFS = true
148 #Most QC compilers translate ‘if(a_vector)’ directly as an IF on
149 #the vector, which means only the x-component is checked. This
150 #option causes vectors to be cast to actual booleans via a NOT_V
151 #and, if necessary, a NOT_F chained to it.
153 # if (a_vector) // becomes
156 # a = a_vector && a_float // becomes
157 # a = !!a_vector && a_float
162 #An empty string is considered to be true everywhere. The NOT_S
163 #instruction usually considers an empty string to be false, this
164 #option effectively causes the unary not in strings to use NOT_F
167 TRUE_EMPTY_STRINGS = false
170 #An empty string is considered to be false everywhere. This means
171 #loops and if statements which depend on a string will perform a
172 #NOT_S instruction on the string before using it.
174 FALSE_EMPTY_STRINGS = true
177 #Enable utf8 characters. This allows utf-8 encoded character con‐
178 #stants, and escape sequence codepoints in the valid utf-8 range.
179 #Effectively enabling escape sequences like '\{x2211}'.
184 #When a warning is treated as an error, and this option is set
185 #(which it is by default), it is like any other error and will
186 #cause compilation to stop. When disabling this flag by using
187 #-fno-bail-on-werror, compilation will continue until the end, but
188 #no output is generated. Instead the first such error message's
191 BAIL_ON_WERROR = false
194 #Allow loops to be labeled, and allow 'break' and 'continue' to
195 #take an optional label to decide which loop to actually jump out
198 # for :outer (i = 0; i < n; ++i) {
209 #Adds a global named 'nil' which is of no type and can be assigned
210 #to anything. No typechecking will be performed on assignments.
211 #Assigning to it is forbidden, using it in any other kind of
212 #expression is also not allowed.
214 #Note that this is different from fteqcc's __NULL__: In fteqcc,
215 #__NULL__ maps to the integer written as '0i'. It's can be
216 #assigned to function pointers and integers, but it'll error about
217 #invalid instructions when assigning it to floats without enabling
218 #the FTE instruction set. There's also a bug which allows it to be
219 #assigned to vectors, for which the source will be the global at
220 #offset 0, meaning the vector's y and z components will contain
221 #the OFS_RETURN x and y components.#
223 #In that gmqcc the nil global is an actual global filled with
224 #zeroes, and can be assigned to anything including fields, vectors
225 #or function pointers, and they end up becoming zeroed.
231 #Various effects, usually to weaken some conditions.
233 # Allow local variables named ‘nil’. (This will not
234 # allow declaring a global of that name.)
239 #Allow variadic parameters to be accessed by QC code. This can be
240 #achieved via the '...' function, which takes a parameter index
245 # void vafunc(string...count) {
247 # for (i = 0; i < count; ++i)
248 # print(...(i, string), "\n");
254 #Most Quake VMs, including the one from FTEQW or up till recently
255 #Darkplaces, do not cope well with vector instructions with over‐
256 #lapping input and output. This option will avoid producing such
259 LEGACY_VECTOR_MATHS = false
262 #Usually builtin-numbers are just immediate constants. With this
263 #flag expressions can be used, as long as they are compile-time
268 # void printA() = #1; // the usual way
269 # void printB() = #2-1; // with a constant expression
271 EXPRESSIONS_FOR_BUILTINS = true
274 #Enabiling this option will allow assigning values or expressions
275 #to the return keyword as if it were a local variable of the same
276 #type as the function's signature's return type.
280 # float bar() { return 1024; }
283 # return; // returns value of bar (this can be omitted)
286 RETURN_ASSIGNMENTS = true
289 #When passing on varargs to a different functions, this turns some
290 #static error cases into warnings. Like when the caller's varargs
291 #are restricted to a different type than the callee's parameter.
292 #Or a list of unrestricted varargs is passed into restricted
295 UNSAFE_VARARGS = false
298 #Always use STORE_F, LOAD_F, STOREP_F when accessing scalar variables.
299 #This is somewhat incorrect assembly instruction use, but in all engines
300 #they do exactly the same. This makes disassembly output harder to read,
301 #breaks decompilers, but causes the output file to be better compressible.
303 TYPELESS_STORES = false
306 #In commutative instructions, always put the lower-numbered operand first.
307 #This shaves off 1 byte of entropy from all these instructions, reducing
308 #compressed size of the output file.
310 SORT_OPERANDS = false
315 #Generate a warning about variables which are declared but never
316 #used. This can be avoided by adding the ‘noref’ keyword in front
317 #of the variable declaration. Additionally a complete section of
318 #unreferenced variables can be opened using ‘#pragma noref 1’ and
319 #closed via ‘#pragma noref 0’.
321 UNUSED_VARIABLE = false
324 #Generate a warning if it is possible that a variable can be used
325 #without prior initialization. Note that this warning is not nec‐
326 #essarily reliable if the initialization happens only under cer‐
327 #tain conditions. The other way is not possible: that the warning
328 #is not generated when uninitialized use is possible.
330 USED_UNINITIALIZED = false
333 #Generate an error when an unrecognized control sequence in a
334 #string is used. Meaning: when there's a character after a back‐
335 #slash in a string which has no known meaning.
337 UNKNOWN_CONTROL_SEQUENCE = false
340 #Warn when using special extensions which are not part of the
346 #Generally QC compilers ignore redeclaration of fields. Here you
347 #can optionally enable a warning.
349 FIELD_REDECLARED = false
352 #Functions which aren't of type void will warn if it possible to
353 #reach the end without returning an actual value.
355 MISSING_RETURN_VALUES = false
358 #Warn about a function call with an invalid number of parameters.
360 INVALID_PARAMETER_COUNT = false
363 #Warn when a locally declared variable shadows variable.
365 LOCAL_SHADOWS = false
368 #Warn when the initialization of a local variable turns the vari‐
369 #able into a constant. This is default behaviour unless
370 #-finitialized-nonconstants is used.
372 LOCAL_CONSTANTS = false
375 #There are only 2 known global variables of type void:
376 #‘end_sys_globals’ and ‘end_sys_fields’. Any other void-variable
379 VOID_VARIABLES = false
382 #A global function which is not declared with the ‘var’ keyword is
383 #expected to have an implementing body, or be a builtin. If nei‐
384 #ther is the case, it implicitly becomes a function pointer, and a
385 #warning is generated.
387 IMPLICIT_FUNCTION_POINTER = false
390 #Currently there's no way for an in QC implemented function to
391 #access variadic parameters. If a function with variadic parame‐
392 #ters has an implementing body, a warning will be generated.
394 VARIADIC_FUNCTION = false
397 #Generate warnings about ‘$frame’ commands, for instance about
398 #duplicate frame definitions.
403 #Warn about statements which have no effect. Any expression which
404 #does not call a function or assigns a variable.
406 EFFECTLESS_STATEMENT = false
409 #The ‘end_sys_fields’ variable is supposed to be a global variable
410 #of type void. It is also recognized as a field but this will
413 END_SYS_FIELDS = false
416 #Warn when assigning to a function pointer with an unmatching sig‐
417 #nature. This usually happens in cases like assigning the null
418 #function to an entity's .think function pointer.
420 ASSIGN_FUNCTION_TYPES = false
423 #Show warnings created using the preprocessor's '#warning' directive
428 #Warn if there's a preprocessor #if spanning across several files.
433 #Warn about multiple declarations of globals. This seems pretty
434 #common in QC code so you probably do not want this unless you
435 #want to clean up your code.
437 DOUBLE_DECLARATION = false
440 #The combination of const and var is not illegal, however differ‐
441 #ent compilers may handle them differently. We were told, the
442 #intention is to create a function-pointer which is not assigna‐
443 #ble. This is exactly how we interpret it. However for this
444 #interpretation the ‘var’ keyword is considered superfluous (and
445 #philosophically wrong), so it is possible to generate a warning
451 #Warn about multibyte character constants, they do not work right
454 MULTIBYTE_CHARACTER = false
457 #Warn if a ternary expression which contains a comma operator is
458 #used without enclosing parenthesis, since this is most likely not
459 #what you actually want. We recommend the -fcorrect-ternary
462 TERNARY_PRECEDENCE = false
465 #Warn when encountering an unrecognized ‘#pragma’ line.
467 UNKNOWN_PRAGMAS = true
470 #Warn about unreachable code. That is: code after a return state‐
471 #ment, or code after a call to a function marked as 'noreturn'.
473 UNREACHABLE_CODE = true
476 #Enable some warnings added in order to help debugging in the com‐
477 #piler. You won't need this.
482 #Warn on an unknown attribute. The warning will inlclude only the
483 #first token inside the enclosing attribute-brackets. This may
484 #change when the actual attribute syntax is better defined.
486 UNKNOWN_ATTRIBUTE = true
489 #Warn when using reserved names such as ‘nil’.
491 RESERVED_NAMES = true
494 #Warn about global constants (using the ‘const’ keyword) with no
497 UNINITIALIZED_CONSTANT = true
500 #Warn about global variables with no initializing value. This is
501 #off by default, and is added mostly to help find null-values
502 #which are supposed to be replaced by the untyped 'nil' constant.
504 UNINITIALIZED_GLOBAL = true
507 #Warn when a variables is redeclared with a different qualifier.
508 #For example when redeclaring a variable as 'var' which was previ‐
509 #ously marked 'const'.
511 DIFFERENT_QUALIFIERS = true
514 #Similar to the above but for attributes like ‘[[noreturn]]’.
516 DIFFERENT_ATTRIBUTES = true
519 #Warn when a function is marked with the attribute "[[depre‐
520 #cated]]". This flag enables a warning on calls to functions
526 #Warn about possible mistakes caused by missing or wrong parenthe‐
527 #sis, like an assignment in an 'if' condition when there's no
528 #additional set of parens around the assignment.
533 #When passing variadic parameters via ...(N) it can happen that
534 #incompatible types are passed to functions. This enables several
535 #warnings when static typechecking cannot guarantee consistent
541 #When compiling original id1 QC there is a definition for `break`
542 #which conflicts with the 'break' keyword in GMQCC. Enabling this
543 #print a warning when the definition occurs. The definition is
544 #ignored for both cases.
549 #When compiling original QuakeWorld QC there are instances where
550 #code overwrites constants. This is considered an error, however
551 #for QuakeWorld to compile it needs to be treated as a warning
552 #instead, as such this warning only works when -std=qcc.
554 CONST_OVERWRITE = true
557 #Warn about the use of preprocessor directives inside macros.
559 DIRECTIVE_INMACRO = true
562 #When using a function that is not explicitly defined, the compiler
563 #will search its intrinsics table for something that matches that
564 #function name by appending "__builtin_" to it. This behaviour may
565 #be unexpected, so enabling this will produce a diagnostic when
566 #such a function is resolved to a builtin.
572 #Some general peephole optimizations. For instance the code `a = b
573 #+ c` typically generates 2 instructions, an ADD and a STORE. This
574 #optimization removes the STORE and lets the ADD write directly
580 #Tail recursive function calls will be turned into loops to avoid
581 #the overhead of the CALL and RETURN instructions.
583 TAIL_RECURSION = true
586 #Make all functions which use neither local arrays nor have locals
587 #which are seen as possibly uninitialized use the same local sec‐
588 #tion. This should be pretty safe compared to other compilers
589 #which do not check for uninitialized values properly. The problem
590 #is that there's QC code out there which really doesn't initialize
591 #some values. This is fine as long as this kind of optimization
592 #isn't used, but also, only as long as the functions cannot be
593 #called in a recursive manner. Since it's hard to know whether or
594 #not an array is actually fully initialized, especially when ini‐
595 #tializing it via a loop, we assume functions with arrays to be
596 #too dangerous for this optimization.
598 OVERLAP_LOCALS = true
601 #This promotes locally declared variables to "temps". Meaning when
602 #a temporary result of an operation has to be stored somewhere, a
603 #local variable which is not 'alive' at that point can be used to
604 #keep the result. This can reduce the size of the global section.
605 #This will not have declared variables overlap, even if it was
611 #Causes temporary values which do not need to be backed up on a
612 #CALL to not be stored in the function's locals-area. With this, a
613 #CALL to a function may need to back up fewer values and thus exe‐
619 #Don't generate defs for immediate values or even declared con‐
620 #stants. Meaning variables which are implicitly constant or qual‐
621 #ified as such using the 'const' keyword.
623 STRIP_CONSTANT_NAMES = true
626 #Aggressively reuse strings in the string section. When a string
627 #should be added which is the trailing substring of an already
628 #existing string, the existing string's tail will be returned
629 #instead of the new string being added.
631 #For example the following code will only generate 1 string:
633 # print("Hello you!\n");
634 # print("you!\n"); // trailing substring of "Hello you!\n"
636 #There's however one limitation. Strings are still processed in
637 #order, so if the above print statements were reversed, this opti‐
638 #mization would not happen.
640 OVERLAP_STRINGS = true
643 #By default, all parameters of a CALL are copied into the parame‐
644 #ter-globals right before the CALL instructions. This is the easi‐
645 #est and safest way to translate calls, but also adds a lot of
646 #unnecessary copying and unnecessary temporary values. This opti‐
647 #mization makes operations which are used as a parameter evaluate
648 #directly into the parameter-global if that is possible, which is
649 #when there's no other CALL instruction in between.
654 #Usually an empty RETURN instruction is added to the end of a void
655 #typed function. However, additionally after every function a DONE
656 #instruction is added for several reasons. (For example the qcvm's
657 #disassemble switch uses it to know when the function ends.). This
658 #optimization replaces that last RETURN with DONE rather than
659 #adding the DONE additionally.
664 #Because traditional QC code doesn't allow you to access individ‐
665 #ual vector components of a computed vector without storing it in
666 #a local first, sometimes people multiply it by a constant like
667 #‘'0 1 0'’ to get, in this case, the y component of a vector. This
668 #optimization will turn such a multiplication into a direct compo‐
669 #nent access. If the factor is anything other than 1, a float-mul‐
670 #tiplication will be added, which is still faster than a vector
673 VECTOR_COMPONENTS = true
676 #For constant expressions that result in dead code (such as a
677 #branch whos condition can be evaluated at compile-time), this
678 #will eliminate the branch and else body (if present) to produce
681 CONST_FOLD_DCE = true
684 #For constant expressions we can fold them to immediate values.
685 #this option cannot be disabled or enabled, the compiler forces
686 #it to stay enabled by ignoring the value entierly. There are
687 #plans to enable some level of constant fold disabling, but right
688 #now the language can't function without it. This is merley here
689 #as an exercise to the reader.