Gold Fish 2

home *** CD-ROM | disk | FTP | other *** search

/ Gold Fish 2 / goldfish_vol2_cd1.bin / gnu / info / make.info-6 (.txt) < prev next >

Wrap

GNU Info File | 1994-11-12 | 49KB | 877 lines

This is Info file make.info, produced by Makeinfo-1.55 from the input file ./make.texinfo. This file documents the GNU Make utility, which determines automatically which pieces of a large program need to be recompiled, and issues the commands to recompile them. This is Edition 0.47, last updated 1 November 1994, of `The GNU Make Manual', for `make', Version 3.72 Beta. Copyright (C) 1988, '89, '90, '91, '92, '93, '94 Free Software Foundation, Inc. Permission is granted to make and distribute verbatim copies of this manual provided the copyright notice and this permission notice are preserved on all copies. Permission is granted to copy and distribute modified versions of this manual under the conditions for verbatim copying, provided that the entire resulting derived work is distributed under the terms of a permission notice identical to this one. Permission is granted to copy and distribute translations of this manual into another language, under the above conditions for modified versions, except that this permission notice may be stated in a translation approved by the Free Software Foundation. File: make.info, Node: Match-Anything Rules, Next: Canceling Rules, Prev: Pattern Match, Up: Pattern Rules Match-Anything Pattern Rules ---------------------------- When a pattern rule's target is just `%', it matches any file name whatever. We call these rules "match-anything" rules. They are very useful, but it can take a lot of time for `make' to think about them, because it must consider every such rule for each file name listed either as a target or as a dependency. Suppose the makefile mentions `foo.c'. For this target, `make' would have to consider making it by linking an object file `foo.c.o', or by C compilation-and-linking in one step from `foo.c.c', or by Pascal compilation-and-linking from `foo.c.p', and many other possibilities. We know these possibilities are ridiculous since `foo.c' is a C source file, not an executable. If `make' did consider these possibilities, it would ultimately reject them, because files such as `foo.c.o' and `foo.c.p' would not exist. But these possibilities are so numerous that `make' would run very slowly if it had to consider them. To gain speed, we have put various constraints on the way `make' considers match-anything rules. There are two different constraints that can be applied, and each time you define a match-anything rule you must choose one or the other for that rule. One choice is to mark the match-anything rule as "terminal" by defining it with a double colon. When a rule is terminal, it does not apply unless its dependencies actually exist. Dependencies that could be made with other implicit rules are not good enough. In other words, no further chaining is allowed beyond a terminal rule. For example, the built-in implicit rules for extracting sources from RCS and SCCS files are terminal; as a result, if the file `foo.c,v' does not exist, `make' will not even consider trying to make it as an intermediate file from `foo.c,v.o' or from `RCS/SCCS/s.foo.c,v'. RCS and SCCS files are generally ultimate source files, which should not be remade from any other files; therefore, `make' can save time by not looking for ways to remake them. If you do not mark the match-anything rule as terminal, then it is nonterminal. A nonterminal match-anything rule cannot apply to a file name that indicates a specific type of data. A file name indicates a specific type of data if some non-match-anything implicit rule target matches it. For example, the file name `foo.c' matches the target for the pattern rule `%.c : %.y' (the rule to run Yacc). Regardless of whether this rule is actually applicable (which happens only if there is a file `foo.y'), the fact that its target matches is enough to prevent consideration of any nonterminal match-anything rules for the file `foo.c'. Thus, `make' will not even consider trying to make `foo.c' as an executable file from `foo.c.o', `foo.c.c', `foo.c.p', etc. The motivation for this constraint is that nonterminal match-anything rules are used for making files containing specific types of data (such as executable files) and a file name with a recognized suffix indicates some other specific type of data (such as a C source file). Special built-in dummy pattern rules are provided solely to recognize certain file names so that nonterminal match-anything rules will not be considered. These dummy rules have no dependencies and no commands, and they are ignored for all other purposes. For example, the built-in implicit rule %.p : exists to make sure that Pascal source files such as `foo.p' match a specific target pattern and thereby prevent time from being wasted looking for `foo.p.o' or `foo.p.c'. Dummy pattern rules such as the one for `%.p' are made for every suffix listed as valid for use in suffix rules (*note Old-Fashioned Suffix Rules: Suffix Rules.). File: make.info, Node: Canceling Rules, Prev: Match-Anything Rules, Up: Pattern Rules Canceling Implicit Rules ------------------------ You can override a built-in implicit rule (or one you have defined yourself) by defining a new pattern rule with the same target and dependencies, but different commands. When the new rule is defined, the built-in one is replaced. The new rule's position in the sequence of implicit rules is determined by where you write the new rule. You can cancel a built-in implicit rule by defining a pattern rule with the same target and dependencies, but no commands. For example, the following would cancel the rule that runs the assembler: %.o : %.s File: make.info, Node: Last Resort, Next: Suffix Rules, Prev: Pattern Rules, Up: Implicit Rules Defining Last-Resort Default Rules ================================== You can define a last-resort implicit rule by writing a terminal match-anything pattern rule with no dependencies (*note Match-Anything Rules::.). This is just like any other pattern rule; the only thing special about it is that it will match any target. So such a rule's commands are used for all targets and dependencies that have no commands of their own and for which no other implicit rule applies. For example, when testing a makefile, you might not care if the source files contain real data, only that they exist. Then you might do this: %:: touch $@ to cause all the source files needed (as dependencies) to be created automatically. You can instead define commands to be used for targets for which there are no rules at all, even ones which don't specify commands. You do this by writing a rule for the target `.DEFAULT'. Such a rule's commands are used for all dependencies which do not appear as targets in any explicit rule, and for which no implicit rule applies. Naturally, there is no `.DEFAULT' rule unless you write one. If you use `.DEFAULT' with no commands or dependencies: .DEFAULT: the commands previously stored for `.DEFAULT' are cleared. Then `make' acts as if you had never defined `.DEFAULT' at all. If you do not want a target to get the commands from a match-anything pattern rule or `.DEFAULT', but you also do not want any commands to be run for the target, you can give it empty commands (*note Defining Empty Commands: Empty Commands.). You can use a last-resort rule to override part of another makefile. *Note Overriding Part of Another Makefile: Overriding Makefiles. File: make.info, Node: Suffix Rules, Next: Search Algorithm, Prev: Last Resort, Up: Implicit Rules Old-Fashioned Suffix Rules ========================== "Suffix rules" are the old-fashioned way of defining implicit rules for `make'. Suffix rules are obsolete because pattern rules are more general and clearer. They are supported in GNU `make' for compatibility with old makefiles. They come in two kinds: "double-suffix" and "single-suffix". A double-suffix rule is defined by a pair of suffixes: the target suffix and the source suffix. It matches any file whose name ends with the target suffix. The corresponding implicit dependency is made by replacing the target suffix with the source suffix in the file name. A two-suffix rule whose target and source suffixes are `.o' and `.c' is equivalent to the pattern rule `%.o : %.c'. A single-suffix rule is defined by a single suffix, which is the source suffix. It matches any file name, and the corresponding implicit dependency name is made by appending the source suffix. A single-suffix rule whose source suffix is `.c' is equivalent to the pattern rule `% : %.c'. Suffix rule definitions are recognized by comparing each rule's target against a defined list of known suffixes. When `make' sees a rule whose target is a known suffix, this rule is considered a single-suffix rule. When `make' sees a rule whose target is two known suffixes concatenated, this rule is taken as a double-suffix rule. For example, `.c' and `.o' are both on the default list of known suffixes. Therefore, if you define a rule whose target is `.c.o', `make' takes it to be a double-suffix rule with source suffix `.c' and target suffix `.o'. Here is the old-fashioned way to define the rule for compiling a C source file: .c.o: $(CC) -c $(CFLAGS) $(CPPFLAGS) -o $@ $< Suffix rules cannot have any dependencies of their own. If they have any, they are treated as normal files with funny names, not as suffix rules. Thus, the rule: .c.o: foo.h $(CC) -c $(CFLAGS) $(CPPFLAGS) -o $@ $< tells how to make the file `.c.o' from the dependency file `foo.h', and is not at all like the pattern rule: %.o: %.c foo.h $(CC) -c $(CFLAGS) $(CPPFLAGS) -o $@ $< which tells how to make `.o' files from `.c' files, and makes all `.o' files using this pattern rule also depend on `foo.h'. Suffix rules with no commands are also meaningless. They do not remove previous rules as do pattern rules with no commands (*note Canceling Implicit Rules: Canceling Rules.). They simply enter the suffix or pair of suffixes concatenated as a target in the data base. The known suffixes are simply the names of the dependencies of the special target `.SUFFIXES'. You can add your own suffixes by writing a rule for `.SUFFIXES' that adds more dependencies, as in: .SUFFIXES: .hack .win which adds `.hack' and `.win' to the end of the list of suffixes. If you wish to eliminate the default known suffixes instead of just adding to them, write a rule for `.SUFFIXES' with no dependencies. By special dispensation, this eliminates all existing dependencies of `.SUFFIXES'. You can then write another rule to add the suffixes you want. For example, .SUFFIXES: # Delete the default suffixes .SUFFIXES: .c .o .h # Define our suffix list The `-r' or `--no-builtin-rules' flag causes the default list of suffixes to be empty. The variable `SUFFIXES' is defined to the default list of suffixes before `make' reads any makefiles. You can change the list of suffixes with a rule for the special target `.SUFFIXES', but that does not alter this variable. File: make.info, Node: Search Algorithm, Prev: Suffix Rules, Up: Implicit Rules Implicit Rule Search Algorithm ============================== Here is the procedure `make' uses for searching for an implicit rule for a target T. This procedure is followed for each double-colon rule with no commands, for each target of ordinary rules none of which have commands, and for each dependency that is not the target of any rule. It is also followed recursively for dependencies that come from implicit rules, in the search for a chain of rules. Suffix rules are not mentioned in this algorithm because suffix rules are converted to equivalent pattern rules once the makefiles have been read in. For an archive member target of the form `ARCHIVE(MEMBER)', the following algorithm is run twice, first using the entire target name T, and second using `(MEMBER)' as the target T if the first run found no rule. 1. Split T into a directory part, called D, and the rest, called N. For example, if T is `src/foo.o', then D is `src/' and N is `foo.o'. 2. Make a list of all the pattern rules one of whose targets matches T or N. If the target pattern contains a slash, it is matched against T; otherwise, against N. 3. If any rule in that list is *not* a match-anything rule, then remove all nonterminal match-anything rules from the list. 4. Remove from the list all rules with no commands. 5. For each pattern rule in the list: a. Find the stem S, which is the nonempty part of T or N matched by the `%' in the target pattern. b. Compute the dependency names by substituting S for `%'; if the target pattern does not contain a slash, append D to the front of each dependency name. c. Test whether all the dependencies exist or ought to exist. (If a file name is mentioned in the makefile as a target or as an explicit dependency, then we say it ought to exist.) If all dependencies exist or ought to exist, or there are no dependencies, then this rule applies. 6. If no pattern rule has been found so far, try harder. For each pattern rule in the list: a. If the rule is terminal, ignore it and go on to the next rule. b. Compute the dependency names as before. c. Test whether all the dependencies exist or ought to exist. d. For each dependency that does not exist, follow this algorithm recursively to see if the dependency can be made by an implicit rule. e. If all dependencies exist, ought to exist, or can be made by implicit rules, then this rule applies. 7. If no implicit rule applies, the rule for `.DEFAULT', if any, applies. In that case, give T the same commands that `.DEFAULT' has. Otherwise, there are no commands for T. Once a rule that applies has been found, for each target pattern of the rule other than the one that matched T or N, the `%' in the pattern is replaced with S and the resultant file name is stored until the commands to remake the target file T are executed. After these commands are executed, each of these stored file names are entered into the data base and marked as having been updated and having the same update status as the file T. When the commands of a pattern rule are executed for T, the automatic variables are set corresponding to the target and dependencies. *Note Automatic Variables: Automatic. File: make.info, Node: Archives, Next: Features, Prev: Implicit Rules, Up: Top Using `make' to Update Archive Files ************************************ "Archive files" are files containing named subfiles called "members"; they are maintained with the program `ar' and their main use is as subroutine libraries for linking. * Menu: * Archive Members:: Archive members as targets. * Archive Update:: The implicit rule for archive member targets. * Archive Pitfalls:: Dangers to watch out for when using archives. * Archive Suffix Rules:: You can write a special kind of suffix rule for updating archives. File: make.info, Node: Archive Members, Next: Archive Update, Up: Archives Archive Members as Targets ========================== An individual member of an archive file can be used as a target or dependency in `make'. You specify the member named MEMBER in archive file ARCHIVE as follows: ARCHIVE(MEMBER) This construct is available only in targets and dependencies, not in commands! Most programs that you might use in commands do not support this syntax and cannot act directly on archive members. Only `ar' and other programs specifically designed to operate on archives can do so. Therefore, valid commands to update an archive member target probably must use `ar'. For example, this rule says to create a member `hack.o' in archive `foolib' by copying the file `hack.o': foolib(hack.o) : hack.o ar cr foolib hack.o In fact, nearly all archive member targets are updated in just this way and there is an implicit rule to do it for you. *Note:* The `c' flag to `ar' is required if the archive file does not already exist. To specify several members in the same archive, you can write all the member names together between the parentheses. For example: foolib(hack.o kludge.o) is equivalent to: foolib(hack.o) foolib(kludge.o) You can also use shell-style wildcards in an archive member reference. *Note Using Wildcard Characters in File Names: Wildcards. For example, `foolib(*.o)' expands to all existing members of the `foolib' archive whose names end in `.o'; perhaps `foolib(hack.o) foolib(kludge.o)'. File: make.info, Node: Archive Update, Next: Archive Pitfalls, Prev: Archive Members, Up: Archives Implicit Rule for Archive Member Targets ======================================== Recall that a target that looks like `A(M)' stands for the member named M in the archive file A. When `make' looks for an implicit rule for such a target, as a special feature it considers implicit rules that match `(M)', as well as those that match the actual target `A(M)'. This causes one special rule whose target is `(%)' to match. This rule updates the target `A(M)' by copying the file M into the archive. For example, it will update the archive member target `foo.a(bar.o)' by copying the *file* `bar.o' into the archive `foo.a' as a *member* named `bar.o'. When this rule is chained with others, the result is very powerful. Thus, `make "foo.a(bar.o)"' (the quotes are needed to protect the `(' and `)' from being interpreted specially by the shell) in the presence of a file `bar.c' is enough to cause the following commands to be run, even without a makefile: cc -c bar.c -o bar.o ar r foo.a bar.o rm -f bar.o Here `make' has envisioned the file `bar.o' as an intermediate file. *Note Chains of Implicit Rules: Chained Rules. Implicit rules such as this one are written using the automatic variable `$%'. *Note Automatic Variables: Automatic. An archive member name in an archive cannot contain a directory name, but it may be useful in a makefile to pretend that it does. If you write an archive member target `foo.a(dir/file.o)', `make' will perform automatic updating with this command: ar r foo.a dir/file.o which has the effect of copying the file `dir/file.o' into a member named `file.o'. In connection with such usage, the automatic variables `%D' and `%F' may be useful. * Menu: * Archive Symbols:: How to update archive symbol directories. File: make.info, Node: Archive Symbols, Up: Archive Update Updating Archive Symbol Directories ----------------------------------- An archive file that is used as a library usually contains a special member named `__.SYMDEF' that contains a directory of the external symbol names defined by all the other members. After you update any other members, you need to update `__.SYMDEF' so that it will summarize the other members properly. This is done by running the `ranlib' program: ranlib ARCHIVEFILE Normally you would put this command in the rule for the archive file, and make all the members of the archive file dependencies of that rule. For example, libfoo.a: libfoo.a(x.o) libfoo.a(y.o) ... ranlib libfoo.a The effect of this is to update archive members `x.o', `y.o', etc., and then update the symbol directory member `__.SYMDEF' by running `ranlib'. The rules for updating the members are not shown here; most likely you can omit them and use the implicit rule which copies files into the archive, as described in the preceding section. This is not necessary when using the GNU `ar' program, which updates the `__.SYMDEF' member automatically. File: make.info, Node: Archive Pitfalls, Next: Archive Suffix Rules, Prev: Archive Update, Up: Archives Dangers When Using Archives =========================== It is important to be careful when using parallel execution (the `-j' switch; *note Parallel Execution: Parallel.) and archives. If multiple `ar' commands run at the same time on the same archive file, they will not know about each other and can corrupt the file. Possibly a future version of `make' will provide a mechanism to circumvent this problem by serializing all commands that operate on the same archive file. But for the time being, you must either write your makefiles to avoid this problem in some other way, or not use `-j'. File: make.info, Node: Archive Suffix Rules, Prev: Archive Pitfalls, Up: Archives Suffix Rules for Archive Files ============================== You can write a special kind of suffix rule for dealing with archive files. *Note Suffix Rules::, for a full explanation of suffix rules. Archive suffix rules are obsolete in GNU `make', because pattern rules for archives are a more general mechanism (*note Archive Update::.). But they are retained for compatibility with other `make's. To write a suffix rule for archives, you simply write a suffix rule using the target suffix `.a' (the usual suffix for archive files). For example, here is the old-fashioned suffix rule to update a library archive from C source files: .c.a: $(CC) $(CFLAGS) $(CPPFLAGS) -c $< -o $*.o $(AR) r $@ $*.o $(RM) $*.o This works just as if you had written the pattern rule: (%.o): %.c $(CC) $(CFLAGS) $(CPPFLAGS) -c $< -o $*.o $(AR) r $@ $*.o $(RM) $*.o In fact, this is just what `make' does when it sees a suffix rule with `.a' as the target suffix. Any double-suffix rule `.X.a' is converted to a pattern rule with the target pattern `(%.o)' and a dependency pattern of `%.X'. Since you might want to use `.a' as the suffix for some other kind of file, `make' also converts archive suffix rules to pattern rules in the normal way (*note Suffix Rules::.). Thus a double-suffix rule `.X.a' produces two pattern rules: `(%.o): %.X' and `%.a: %.X'. File: make.info, Node: Features, Next: Missing, Prev: Archives, Up: Top Features of GNU `make' ********************** Here is a summary of the features of GNU `make', for comparison with and credit to other versions of `make'. We consider the features of `make' in 4.2 BSD systems as a baseline. If you are concerned with writing portable makefiles, you should use only the features of `make' *not* listed here or in *Note Missing::. Many features come from the version of `make' in System V. * The `VPATH' variable and its special meaning. *Note Searching Directories for Dependencies: Directory Search. This feature exists in System V `make', but is undocumented. It is documented in 4.3 BSD `make' (which says it mimics System V's `VPATH' feature). * Included makefiles. *Note Including Other Makefiles: Include. Allowing multiple files to be included with a single directive is a GNU extension. * Variables are read from and communicated via the environment. *Note Variables from the Environment: Environment. * Options passed through the variable `MAKEFLAGS' to recursive invocations of `make'. *Note Communicating Options to a Sub-`make': Options/Recursion. * The automatic variable `$%' is set to the member name in an archive reference. *Note Automatic Variables: Automatic. * The automatic variables `$@', `$*', `$<', `$%', and `$?' have corresponding forms like `$(@F)' and `$(@D)'. We have generalized this to `$^' as an obvious extension. *Note Automatic Variables: Automatic. * Substitution variable references. *Note Basics of Variable References: Reference. * The command-line options `-b' and `-m', accepted and ignored. In System V `make', these options actually do something. * Execution of recursive commands to run `make' via the variable `MAKE' even if `-n', `-q' or `-t' is specified. *Note Recursive Use of `make': Recursion. * Support for suffix `.a' in suffix rules. *Note Archive Suffix Rules::. This feature is obsolete in GNU `make', because the general feature of rule chaining (*note Chains of Implicit Rules: Chained Rules.) allows one pattern rule for installing members in an archive (*note Archive Update::.) to be sufficient. * The arrangement of lines and backslash-newline combinations in commands is retained when the commands are printed, so they appear as they do in the makefile, except for the stripping of initial whitespace. The following features were inspired by various other versions of `make'. In some cases it is unclear exactly which versions inspired which others. * Pattern rules using `%'. This has been implemented in several versions of `make'. We're not sure who invented it first, but it's been spread around a bit. *Note Defining and Redefining Pattern Rules: Pattern Rules. * Rule chaining and implicit intermediate files. This was implemented by Stu Feldman in his version of `make' for AT&T Eighth Edition Research Unix, and later by Andrew Hume of AT&T Bell Labs in his `mk' program (where he terms it "transitive closure"). We do not really know if we got this from either of them or thought it up ourselves at the same time. *Note Chains of Implicit Rules: Chained Rules. * The automatic variable `$^' containing a list of all dependencies of the current target. We did not invent this, but we have no idea who did. *Note Automatic Variables: Automatic. The automatic variable `$+' is a simple extension of `$^'. * The "what if" flag (`-W' in GNU `make') was (as far as we know) invented by Andrew Hume in `mk'. *Note Instead of Executing the Commands: Instead of Execution. * The concept of doing several things at once (parallelism) exists in many incarnations of `make' and similar programs, though not in the System V or BSD implementations. *Note Command Execution: Execution. * Modified variable references using pattern substitution come from SunOS 4. *Note Basics of Variable References: Reference. This functionality was provided in GNU `make' by the `patsubst' function before the alternate syntax was implemented for compatibility with SunOS 4. It is not altogether clear who inspired whom, since GNU `make' had `patsubst' before SunOS 4 was released. * The special significance of `+' characters preceding command lines (*note Instead of Executing the Commands: Instead of Execution.) is mandated by `IEEE Standard 1003.2-1992' (POSIX.2). * The `+=' syntax to append to the value of a variable comes from SunOS 4 `make'. *Note Appending More Text to Variables: Appending. * The syntax `ARCHIVE(MEM1 MEM2...)' to list multiple members in a single archive file comes from SunOS 4 `make'. *Note Archive Members::. * The `-include' directive to include makefiles with no error for a nonexistent file comes from SunOS 4 `make'. (But note that SunOS 4 `make' does not allow multiple makefiles to be specified in one `-include' directive.) The remaining features are inventions new in GNU `make': * Use the `-v' or `--version' option to print version and copyright information. * Use the `-h' or `--help' option to summarize the options to `make'. * Simply-expanded variables. *Note The Two Flavors of Variables: Flavors. * Pass command-line variable assignments automatically through the variable `MAKE' to recursive `make' invocations. *Note Recursive Use of `make': Recursion. * Use the `-C' or `--directory' command option to change directory. *Note Summary of Options: Options Summary. * Make verbatim variable definitions with `define'. *Note Defining Variables Verbatim: Defining. * Declare phony targets with the special target `.PHONY'. Andrew Hume of AT&T Bell Labs implemented a similar feature with a different syntax in his `mk' program. This seems to be a case of parallel discovery. *Note Phony Targets: Phony Targets. * Manipulate text by calling functions. *Note Functions for Transforming Text: Functions. * Use the `-o' or `--old-file' option to pretend a file's modification-time is old. *Note Avoiding Recompilation of Some Files: Avoiding Compilation. * Conditional execution. This feature has been implemented numerous times in various versions of `make'; it seems a natural extension derived from the features of the C preprocessor and similar macro languages and is not a revolutionary concept. *Note Conditional Parts of Makefiles: Conditionals. * Specify a search path for included makefiles. *Note Including Other Makefiles: Include. * Specify extra makefiles to read with an environment variable. *Note The Variable `MAKEFILES': MAKEFILES Variable. * Strip leading sequences of `./' from file names, so that `./FILE' and `FILE' are considered to be the same file. * Use a special search method for library dependencies written in the form `-lNAME'. *Note Directory Search for Link Libraries: Libraries/Search. * Allow suffixes for suffix rules (*note Old-Fashioned Suffix Rules: Suffix Rules.) to contain any characters. In other versions of `make', they must begin with `.' and not contain any `/' characters. * Keep track of the current level of `make' recursion using the variable `MAKELEVEL'. *Note Recursive Use of `make': Recursion. * Specify static pattern rules. *Note Static Pattern Rules: Static Pattern. * Provide selective `vpath' search. *Note Searching Directories for Dependencies: Directory Search. * Provide computed variable references. *Note Basics of Variable References: Reference. * Update makefiles. *Note How Makefiles Are Remade: Remaking Makefiles. System V `make' has a very, very limited form of this functionality in that it will check out SCCS files for makefiles. * Various new built-in implicit rules. *Note Catalogue of Implicit Rules: Catalogue of Rules. * The built-in variable `MAKE_VERSION' gives the version number of `make'. File: make.info, Node: Missing, Next: Makefile Conventions, Prev: Features, Up: Top Incompatibilities and Missing Features ************************************** The `make' programs in various other systems support a few features that are not implemented in GNU `make'. The POSIX.2 standard (`IEEE Standard 1003.2-1992') which specifies `make' does not require any of these features. * A target of the form `FILE((ENTRY))' stands for a member of archive file FILE. The member is chosen, not by name, but by being an object file which defines the linker symbol ENTRY. This feature was not put into GNU `make' because of the nonmodularity of putting knowledge into `make' of the internal format of archive file symbol tables. *Note Updating Archive Symbol Directories: Archive Symbols. * Suffixes (used in suffix rules) that end with the character `~' have a special meaning to System V `make'; they refer to the SCCS file that corresponds to the file one would get without the `~'. For example, the suffix rule `.c~.o' would make the file `N.o' from the SCCS file `s.N.c'. For complete coverage, a whole series of such suffix rules is required. *Note Old-Fashioned Suffix Rules: Suffix Rules. In GNU `make', this entire series of cases is handled by two pattern rules for extraction from SCCS, in combination with the general feature of rule chaining. *Note Chains of Implicit Rules: Chained Rules. * In System V `make', the string `$$@' has the strange meaning that, in the dependencies of a rule with multiple targets, it stands for the particular target that is being processed. This is not defined in GNU `make' because `$$' should always stand for an ordinary `$'. It is possible to get this functionality through the use of static pattern rules (*note Static Pattern Rules: Static Pattern.). The System V `make' rule: $(targets): $$@.o lib.a can be replaced with the GNU `make' static pattern rule: $(targets): %: %.o lib.a * In System V and 4.3 BSD `make', files found by `VPATH' search (*note Searching Directories for Dependencies: Directory Search.) have their names changed inside command strings. We feel it is much cleaner to always use automatic variables and thus make this feature obsolete. * In some Unix `make's, the automatic variable `$*' appearing in the dependencies of a rule has the amazingly strange "feature" of expanding to the full name of the *target of that rule*. We cannot imagine what went on in the minds of Unix `make' developers to do this; it is utterly inconsistent with the normal definition of `$*'. * In some Unix `make's, implicit rule search (*note Using Implicit Rules: Implicit Rules.) is apparently done for *all* targets, not just those without commands. This means you can do: foo.o: cc -c foo.c and Unix `make' will intuit that `foo.o' depends on `foo.c'. We feel that such usage is broken. The dependency properties of `make' are well-defined (for GNU `make', at least), and doing such a thing simply does not fit the model. * GNU `make' does not include any built-in implicit rules for compiling or preprocessing EFL programs. If we hear of anyone who is using EFL, we will gladly add them. * It appears that in SVR4 `make', a suffix rule can be specified with no commands, and it is treated as if it had empty commands (*note Empty Commands::.). For example: .c.a: will override the built-in `.c.a' suffix rule. We feel that it is cleaner for a rule without commands to always simply add to the dependency list for the target. The above example can be easily rewritten to get the desired behavior in GNU `make': .c.a: ; * Some versions of `make' invoke the shell with the `-e' flag, except under `-k' (*note Testing the Compilation of a Program: Testing.). The `-e' flag tells the shell to exit as soon as any program it runs returns a nonzero status. We feel it is cleaner to write each shell command line to stand on its own and not require this special treatment. File: make.info, Node: Makefile Conventions, Next: Quick Reference, Prev: Missing, Up: Top Makefile Conventions ******************** This chapter describes conventions for writing the Makefiles for GNU programs. * Menu: * Makefile Basics:: * Utilities in Makefiles:: * Standard Targets:: * Command Variables:: * Directory Variables:: File: make.info, Node: Makefile Basics, Next: Utilities in Makefiles, Up: Makefile Conventions General Conventions for Makefiles ================================= Every Makefile should contain this line: SHELL = /bin/sh to avoid trouble on systems where the `SHELL' variable might be inherited from the environment. (This is never a problem with GNU `make'.) Different `make' programs have incompatible suffix lists and implicit rules, and this sometimes creates confusion or misbehavior. So it is a good idea to set the suffix list explicitly using only the suffixes you need in the particular Makefile, like this: .SUFFIXES: .SUFFIXES: .c .o The first line clears out the suffix list, the second introduces all suffixes which may be subject to implicit rules in this Makefile. Don't assume that `.' is in the path for command execution. When you need to run programs that are a part of your package during the make, please make sure that it uses `./' if the program is built as part of the make or `$(srcdir)/' if the file is an unchanging part of the source code. Without one of these prefixes, the current search path is used. The distinction between `./' and `$(srcdir)/' is important when using the `--srcdir' option to `configure'. A rule of the form: foo.1 : foo.man sedscript sed -e sedscript foo.man > foo.1 will fail when the current directory is not the source directory, because `foo.man' and `sedscript' are not in the current directory. When using GNU `make', relying on `VPATH' to find the source file will work in the case where there is a single dependency file, since the `make' automatic variable `$<' will represent the source file wherever it is. (Many versions of `make' set `$<' only in implicit rules.) A makefile target like foo.o : bar.c $(CC) -I. -I$(srcdir) $(CFLAGS) -c bar.c -o foo.o should instead be written as foo.o : bar.c $(CC) -I. -I$(srcdir) $(CFLAGS) -c $< -o $@ in order to allow `VPATH' to work correctly. When the target has multiple dependencies, using an explicit `$(srcdir)' is the easiest way to make the rule work well. For example, the target above for `foo.1' is best written as: foo.1 : foo.man sedscript sed -e $(srcdir)/sedscript $(srcdir)/foo.man > $@ File: make.info, Node: Utilities in Makefiles, Next: Standard Targets, Prev: Makefile Basics, Up: Makefile Conventions Utilities in Makefiles ====================== Write the Makefile commands (and any shell scripts, such as `configure') to run in `sh', not in `csh'. Don't use any special features of `ksh' or `bash'. The `configure' script and the Makefile rules for building and installation should not use any utilities directly except these: cat cmp cp echo egrep expr grep ln mkdir mv pwd rm rmdir sed test touch Stick to the generally supported options for these programs. For example, don't use `mkdir -p', convenient as it may be, because most systems don't support it. The Makefile rules for building and installation can also use compilers and related programs, but should do so via `make' variables so that the user can substitute alternatives. Here are some of the programs we mean: ar bison cc flex install ld lex make makeinfo ranlib texi2dvi yacc Use the following `make' variables: $(AR) $(BISON) $(CC) $(FLEX) $(INSTALL) $(LD) $(LEX) $(MAKE) $(MAKEINFO) $(RANLIB) $(TEXI2DVI) $(YACC) When you use `ranlib', you should make sure nothing bad happens if the system does not have `ranlib'. Arrange to ignore an error from that command, and print a message before the command to tell the user that failure of the `ranlib' command does not mean a problem. If you use symbolic links, you should implement a fallback for systems that don't have symbolic links. It is ok to use other utilities in Makefile portions (or scripts) intended only for particular systems where you know those utilities to exist. File: make.info, Node: Standard Targets, Next: Command Variables, Prev: Utilities in Makefiles, Up: Makefile Conventions Standard Targets for Users ========================== All GNU programs should have the following targets in their Makefiles: `all' Compile the entire program. This should be the default target. This target need not rebuild any documentation files; Info files should normally be included in the distribution, and DVI files should be made only when explicitly asked for. `install' Compile the program and copy the executables, libraries, and so on to the file names where they should reside for actual use. If there is a simple test to verify that a program is properly installed, this target should run that test. The commands should create all the directories in which files are to be installed, if they don't already exist. This includes the directories specified as the values of the variables `prefix' and `exec_prefix', as well as all subdirectories that are needed. One way to do this is by means of an `installdirs' target as described below. Use `-' before any command for installing a man page, so that `make' will ignore any errors. This is in case there are systems that don't have the Unix man page documentation system installed. The way to install Info files is to copy them into `$(infodir)' with `$(INSTALL_DATA)' (*note Command Variables::.), and then run the `install-info' program if it is present. `install-info' is a script that edits the Info `dir' file to add or update the menu entry for the given Info file; it will be part of the Texinfo package. Here is a sample rule to install an Info file: $(infodir)/foo.info: foo.info # There may be a newer info file in . than in srcdir. -if test -f foo.info; then d=.; \ else d=$(srcdir); fi; \ $(INSTALL_DATA) $$d/foo.info $@; \ # Run install-info only if it exists. # Use `if' instead of just prepending `-' to the # line so we notice real errors from install-info. # We use `$(SHELL) -c' because some shells do not # fail gracefully when there is an unknown command. if $(SHELL) -c 'install-info --version' \ >/dev/null 2>&1; then \ install-info --infodir=$(infodir) $$d/foo.info; \ else true; fi `uninstall' Delete all the installed files that the `install' target would create (but not the noninstalled files such as `make all' would create). `clean' Delete all files from the current directory that are normally created by building the program. Don't delete the files that record the configuration. Also preserve files that could be made by building, but normally aren't because the distribution comes with them. Delete `.dvi' files here if they are not part of the distribution. `distclean' Delete all files from the current directory that are created by configuring or building the program. If you have unpacked the source and built the program without creating any other files, `make distclean' should leave only the files that were in the distribution. `mostlyclean' Like `clean', but may refrain from deleting a few files that people normally don't want to recompile. For example, the `mostlyclean' target for GCC does not delete `libgcc.a', because recompiling it is rarely necessary and takes a lot of time. `realclean' Delete everything from the current directory that can be reconstructed with this Makefile. This typically includes everything deleted by `distclean', plus more: C source files produced by Bison, tags tables, Info files, and so on. One exception, however: `make realclean' should not delete `configure' even if `configure' can be remade using a rule in the Makefile. More generally, `make realclean' should not delete anything that needs to exist in order to run `configure' and then begin to build the program. `TAGS' Update a tags table for this program. `info' Generate any Info files needed. The best way to write the rules is as follows: info: foo.info foo.info: foo.texi chap1.texi chap2.texi $(MAKEINFO) $(srcdir)/foo.texi You must define the variable `MAKEINFO' in the Makefile. It should run the `makeinfo' program, which is part of the Texinfo distribution. `dvi' Generate DVI files for all TeXinfo documentation. For example: dvi: foo.dvi foo.dvi: foo.texi chap1.texi chap2.texi $(TEXI2DVI) $(srcdir)/foo.texi You must define the variable `TEXI2DVI' in the Makefile. It should run the program `texi2dvi', which is part of the Texinfo distribution. Alternatively, write just the dependencies, and allow GNU Make to provide the command. `dist' Create a distribution tar file for this program. The tar file should be set up so that the file names in the tar file start with a subdirectory name which is the name of the package it is a distribution for. This name can include the version number. For example, the distribution tar file of GCC version 1.40 unpacks into a subdirectory named `gcc-1.40'. The easiest way to do this is to create a subdirectory appropriately named, use `ln' or `cp' to install the proper files in it, and then `tar' that subdirectory. The `dist' target should explicitly depend on all non-source files that are in the distribution, to make sure they are up to date in the distribution. *Note Making Releases: (standards)Releases. `check' Perform self-tests (if any). The user must build the program before running the tests, but need not install the program; you should write the self-tests so that they work when the program is built but not installed. The following targets are suggested as conventional names, for programs in which they are useful. `installcheck' Perform installation tests (if any). The user must build and install the program before running the tests. You should not assume that `$(bindir)' is in the search path. `installdirs' It's useful to add a target named `installdirs' to create the directories where files are installed, and their parent directories. There is a script called `mkinstalldirs' which is convenient for this; find it in the Texinfo package.You can use a rule like this: # Make sure all installation directories (e.g. $(bindir)) # actually exist by making them if necessary. installdirs: mkinstalldirs $(srcdir)/mkinstalldirs $(bindir) $(datadir) \ $(libdir) $(infodir) \ $(mandir) File: make.info, Node: Command Variables, Next: Directory Variables, Prev: Standard Targets, Up: Makefile Conventions Variables for Specifying Commands ================================= Makefiles should provide variables for overriding certain commands, options, and so on. In particular, you should run most utility programs via variables. Thus, if you use Bison, have a variable named `BISON' whose default value is set with `BISON = bison', and refer to it with `$(BISON)' whenever you need to use Bison. File management utilities such as `ln', `rm', `mv', and so on, need not be referred to through variables in this way, since users don't need to replace them with other programs. Each program-name variable should come with an options variable that is used to supply options to the program. Append `FLAGS' to the program-name variable name to get the options variable name--for example, `BISONFLAGS'. (The name `CFLAGS' is an exception to this rule, but we keep it because it is standard.) Use `CPPFLAGS' in any compilation command that runs the preprocessor, and use `LDFLAGS' in any compilation command that does linking as well as in any direct use of `ld'. If there are C compiler options that *must* be used for proper compilation of certain files, do not include them in `CFLAGS'. Users expect to be able to specify `CFLAGS' freely themselves. Instead, arrange to pass the necessary options to the C compiler independently of `CFLAGS', by writing them explicitly in the compilation commands or by defining an implicit rule, like this: CFLAGS = -g ALL_CFLAGS = -I. $(CFLAGS) .c.o: $(CC) -c $(CPPFLAGS) $(ALL_CFLAGS) $< Do include the `-g' option in `CFLAGS', because that is not *required* for proper compilation. You can consider it a default that is only recommended. If the package is set up so that it is compiled with GCC by default, then you might as well include `-O' in the default value of `CFLAGS' as well. Put `CFLAGS' last in the compilation command, after other variables containing compiler options, so the user can use `CFLAGS' to override the others. Every Makefile should define the variable `INSTALL', which is the basic command for installing a file into the system. Every Makefile should also define the variables `INSTALL_PROGRAM' and `INSTALL_DATA'. (The default for each of these should be `$(INSTALL)'.) Then it should use those variables as the commands for actual installation, for executables and nonexecutables respectively. Use these variables as follows: $(INSTALL_PROGRAM) foo $(bindir)/foo $(INSTALL_DATA) libfoo.a $(libdir)/libfoo.a Always use a file name, not a directory name, as the second argument of the installation commands. Use a separate command for each file to be installed.