1 ------------------------------------------------------------------------------
3 -- GNAT COMPILER COMPONENTS --
9 -- Copyright (C) 1992-2004 Free Software Foundation, Inc. --
11 -- GNAT is free software; you can redistribute it and/or modify it under --
12 -- terms of the GNU General Public License as published by the Free Soft- --
13 -- ware Foundation; either version 2, or (at your option) any later ver- --
14 -- sion. GNAT is distributed in the hope that it will be useful, but WITH- --
15 -- OUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY --
16 -- or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License --
17 -- for more details. You should have received a copy of the GNU General --
18 -- Public License distributed with GNAT; see file COPYING. If not, write --
19 -- to the Free Software Foundation, 59 Temple Place - Suite 330, Boston, --
20 -- MA 02111-1307, USA. --
22 -- GNAT was originally developed by the GNAT team at New York University. --
23 -- Extensive contributions were provided by Ada Core Technologies Inc. --
25 ------------------------------------------------------------------------------
27 -- This package contains the routines for writing the library information
31 -----------------------------------
32 -- Format of Library Information --
33 -----------------------------------
35 -- This section describes the format of the library information that is
36 -- associated with object files. The exact method of this association is
37 -- potentially implementation dependent and is described and implemented
38 -- in package ali. From the point of view of the description here, all we
39 -- need to know is that the information is represented as a string of
40 -- characters that is somehow associated with an object file, and can be
41 -- retrieved. If no library information exists for a given object file,
42 -- then we take this as equivalent to the non-existence of the object
43 -- file, as if source file has not been previously compiled.
45 -- The library information is written as a series of lines of the form:
47 -- Key_Character parameter parameter ...
49 -- The following sections describe the format of these lines in detail
51 --------------------------------------
52 -- Making Changes to the ALI Format --
53 --------------------------------------
55 -- A number of tools use ali.adb to parse ali files. This means
56 -- that changes to this format can cause old versions of these tools
57 -- to be incompatible with new versions of the compiler. Any changes
58 -- to ali file formats must be carefully evaluated to understand any
59 -- such possible conflicts, and in particular, it is very undesirable
60 -- to create conflicts between older versions of GPS and newer versions
63 -- If the following guidelines are respected, downward compatibility
64 -- problems (old tools reading new ali files) should be minimized:
66 -- The basic key character format must be kept.
68 -- The V line must be the first line, this is checked by ali.adb
69 -- even in Ignore_Errors mode, and is used to verify that the file
70 -- at hand is indeed likely intended to be an ali file.
72 -- The P line must be present, though may be modified in contents
73 -- according to remaining guidelines. Again, ali.adb assumes the
74 -- P line is present even in Ignore_Errors mode.
76 -- New modifiers can generally be added (in particular adding new
77 -- two letter modifiers to the P or U lines is always safe)
79 -- Adding entirely new lines (with a new key letter) to the ali
80 -- file is always safe, at any point (other than before the V
81 -- line), since suchy lines will be ignored.
83 -- Following the guidelines in this section should ensure that this
84 -- problem is minimized and that old tools will be able to deal
85 -- successfully with new ali formats. Note that this does not apply
86 -- to the compiler itself, which always requires consistency between
87 -- the ali files and the binder. That is because one of the main
88 -- functions of the binder is to ensure consistency of the partition,
89 -- and this can be compromised if the ali files are inconsistent.
95 -- The initial header lines in the file give information about the
96 -- compilation environment, and identify other special information
97 -- such as main program parameters.
103 -- V "xxxxxxxxxxxxxxxx"
105 -- This line indicates the library output version, as defined in
106 -- Gnatvsn. It ensures that separate object modules of a program are
107 -- consistent. It has to be changed if anything changes which would
108 -- affect successful binding of separately compiled modules.
109 -- Examples of such changes are modifications in the format of the
110 -- library info described in this package, or modifications to
111 -- calling sequences, or to the way that data is represented.
113 -- Note: the V line absolutely must be the first line, and no change
114 -- to the ALI format should change this, since even in Ignore_Errors
115 -- mode, Scan_ALI insists on finding a V line.
117 -- ---------------------
118 -- -- M Main Program --
119 -- ---------------------
121 -- M type [priority] [T=time-slice] W=?
123 -- This line appears only if the main unit for this file is
124 -- suitable for use as a main program. The parameters are:
128 -- P for a parameterless procedure
129 -- F for a function returning a value of integral type
130 -- (used for writing a main program returning an exit status)
134 -- Present only if there was a valid pragma Priority in the
135 -- corresponding unit to set the main task priority. It is
136 -- an unsigned decimal integer.
140 -- Present only if there was a valid pragma Time_Slice in the
141 -- corresponding unit. It is an unsigned decimal integer in
142 -- the range 0 .. 10**9 giving the time slice value in units
143 -- of milliseconds. The actual significance of this parameter
144 -- is target dependent.
148 -- This parameter indicates the wide character encoding
149 -- method used when compiling the main program file. The ?
150 -- character is the single character used in the -gnatW?
151 -- switch. This is used to provide the default wide-character
152 -- encoding for Wide_Text_IO files.
160 -- One of these lines appears for each of the arguments present
161 -- in the call to the gnat1 program. This can be used if it is
162 -- necessary to reconstruct this call (e.g. for fix and continue)
164 -- -------------------
165 -- -- P Parameters --
166 -- -------------------
170 -- Indicates various information that applies to the compilation
171 -- of the corresponding source unit. Parameters is a sequence of
172 -- zero or more two letter codes that indicate configuration
173 -- pragmas and other parameters that apply:
175 -- The arguments are as follows:
177 -- CE Compilation errors. If this is present it means that the
178 -- ali file resulted from a compilation with the -gnatQ
179 -- switch set, and illegalities were detected. The ali
180 -- file contents may not be completely reliable, but the
181 -- format will be correct and complete. Note that NO is
182 -- always present if CE is present.
184 -- DB Detect_Blocking pragma is in effect for all units in
187 -- FD Configuration pragmas apply to all the units in this
188 -- file specifying a possibly non-standard floating point
189 -- format (VAX float with Long_Float using D_Float)
191 -- FG Configuration pragmas apply to all the units in this
192 -- file specifying a possibly non-standard floating point
193 -- format (VAX float with Long_Float using G_Float)
195 -- FI Configuration pragmas apply to all the units in this
196 -- file specifying a possibly non-standard floating point
197 -- format (IEEE Float)
199 -- Lx A valid Locking_Policy pragma applies to all the units
200 -- in this file, where x is the first character (upper case)
201 -- of the policy name (e.g. 'C' for Ceiling_Locking)
203 -- NO No object. This flag indicates that the units in this
204 -- file were not compiled to produce an object. This can
205 -- occur as a result of the use of -gnatc, or if no object
206 -- can be produced (e.g. when a package spec is compiled
207 -- instead of the body, or a subunit on its own).
209 -- NR No_Run_Time. Indicates that a pragma No_Run_Time applies
210 -- to all units in the file.
212 -- NS Normalize_Scalars pragma in effect for all units in
215 -- Qx A valid Queueing_Policy pragma applies to all the units
216 -- in this file, where x is the first character (upper case)
217 -- of the policy name (e.g. 'P' for Priority_Queueing).
219 -- SL Indicates that the unit is an Interface to a Standalone
220 -- Library. Note that this indication is never given by the
221 -- compiler, but is added by the Project Manager in gnatmake
222 -- when an Interface ALI file is copied to the library
225 -- SS This unit references System.Secondary_Stack (that is,
226 -- the unit makes use of the secondary stack facilities).
228 -- Tx A valid Task_Dispatching_Policy pragma applies to all
229 -- the units in this file, where x is the first character
230 -- (upper case) of the corresponding policy name (e.g. 'F'
231 -- for FIFO_Within_Priorities).
233 -- UA Unreserve_All_Interrupts pragma was processed in one or
234 -- more units in this file
236 -- UX Generated code contains unit exception table pointer
237 -- (i.e. it uses zero-cost exceptions, and there is at
238 -- least one subprogram present).
240 -- ZX Units in this file use zero-cost exceptions and have
241 -- generated exception tables. If ZX is not present, the
242 -- longjmp/setjmp exception scheme is in use.
244 -- Note that language defined units never output policy (Lx,Tx,Qx)
245 -- parameters. Language defined units must correctly handle all
246 -- possible cases. These values are checked for consistency by the
247 -- binder and then copied to the generated binder output file.
249 -- Note: The P line must be present. Even in Ignore_Errors mode,
250 -- Scan_ALI insists on finding a P line. So if changes are made to
251 -- the ALI format, they should not include removing the P line!
253 -- ---------------------
254 -- -- R Restrictions --
255 -- ---------------------
257 -- The first R line records the status of restrictions generated by pragma
258 -- Restrictions encountered, as well as information on what the compiler
259 -- has been able to determine with respect to restrictions violations.
262 -- R <<restriction-characters>> <<restriction-param-id-entries>>
264 -- The first parameter is a string of characters that records
265 -- information regarding restrictions that do not take parameter
266 -- not take parameter values. It is a string of characters, one
267 -- character for each value (in order) in All_Boolean_Restrictions.
268 -- There are three possible settings for each restriction:
270 -- r Restricted. Unit was compiled under control of a pragma
271 -- Restrictions for the corresponding restriction. In
272 -- this case the unit certainly does not violate the
273 -- Restriction, since this would have been detected by
276 -- n Not used. The unit was not compiled under control of a
277 -- pragma Restrictions for the corresponding restriction,
278 -- and does not make any use of the referenced feature.
280 -- v Violated. The unit was not compiled under control of a
281 -- pragma Restrictions for the corresponding restriction,
282 -- and it does indeed use the referenced feature.
284 -- This information is used in the binder to check consistency,
285 -- i.e. to detect cases where one unit has "r" and another unit
286 -- has "v", which is not permitted, since these restrictions
287 -- are partition-wide.
289 -- The second parameter, which immediately follows the first (with
290 -- no separating space) gives restriction information for identifiers
291 -- for which a parameter is given.
293 -- The parameter is a string of entries, one for each value in
294 -- Restrict.All_Parameter_Restrictions. Each entry has two
295 -- components in sequence, the first indicating whether or not
296 -- there is a restriction, and the second indicating whether
297 -- or not the compiler detected violations. In the boolean case
298 -- it is not necessary to separate these, since if a restriction
299 -- is set, and violated, that is an error. But in the parameter
300 -- case, this is not true. For example, we can have a unit with
301 -- a pragma Restrictions (Max_Tasks => 4), where the compiler
302 -- can detect that there are exactly three tasks declared. Both
303 -- of these pieces of information must be passed to the binder.
304 -- The parameter of 4 is important in case the total number of
305 -- tasks in the partition is greater than 4. The parameter of
306 -- 3 is important in case some other unit has a restrictions
307 -- pragma with Max_Tasks=>2.
309 -- The component for the presence of restriction has one of two
312 -- n No pragma for this restriction is present in the
313 -- set of units for this ali file.
315 -- rN At least one pragma for this restriction is present
316 -- in the set of units for this ali file. The value N
317 -- is the minimum parameter value encountered in any
318 -- such pragma. N is in the range of Integer (a value
319 -- larger than N'Last causes the pragma to be ignored).
321 -- The component for the violation detection has one of three
324 -- n No violations were detected by the compiler
326 -- vN A violation was detected. N is either the maximum or total
327 -- count of violations (depending on the checking type) in
328 -- all the units represented by the ali file). Note that
329 -- this setting is only allowed for restrictions that are
330 -- in Checked_[Max|Sum]_Parameter_Restrictions. The value
331 -- here is known to be exact by the compiler and is in the
334 -- vN+ A violation was detected. The compiler cannot determine
335 -- the exact count of violations, but it is at least N.
337 -- There are no spaces within the parameter string, so the entry
338 -- described above in the header of this section for Max_Tasks would
339 -- appear as the string r4v3.
341 -- Note: The restrictions line is required to be present. Even in
342 -- Ignore_Errors mode, Scan_ALI expects to find an R line and will
343 -- signal a fatal error if it is missing. This means that future
344 -- changes to the ALI file format must retain the R line.
346 -- Subsequent R lines are present only if pragma Restriction No_Dependence
347 -- is used. There is one such line for each such pragma appearing in the
348 -- extended main unit. The format is
352 -- Here the unit name is in all lower case. The components of the unit
353 -- name are separated by periods. The names themselves are in encoded
354 -- form, as documented in Namet.
356 -- ------------------------
357 -- -- I Interrupt States --
358 -- ------------------------
360 -- I interrupt-number interrupt-state line-number
362 -- This line records information from an Interrupt_State pragma.
363 -- There is one line for each separate pragma, and if no such
364 -- pragmas are used, then no I lines are present.
366 -- The interrupt-number is an unsigned positive integer giving
367 -- the value of the interrupt as defined in Ada.Interrupts.Names.
369 -- The interrupt-state is one of r/s/u for Runtime/System/User
371 -- The line number is an unsigned decimal integer giving the
372 -- line number of the corresponding Interrupt_State pragma.
373 -- This is used in consistency messages.
375 ----------------------------
376 -- Compilation Unit Lines --
377 ----------------------------
379 -- Following these header lines, a set of information lines appears for
380 -- each compilation unit that appears in the corresponding object file.
381 -- In particular, when a package body or subprogram body is compiled,
382 -- there will be two sets of information, one for the spec and one for
383 -- the body. with the entry for the body appearing first. This is the
384 -- only case in which a single ALI file contains more than one unit (in
385 -- particular note that subunits do *not* count as compilation units for
386 -- this purpose, and generate no library information, since they are
389 -- --------------------
390 -- -- U Unit Header --
391 -- --------------------
393 -- The lines for each compilation unit have the following form.
395 -- U unit-name source-name version <<attributes>>
397 -- This line identifies the unit to which this section of the
398 -- library information file applies. The first three parameters are
399 -- the unit name in internal format, as described in package Uname,
400 -- and the name of the source file containing the unit.
402 -- Version is the version given as eight hexadecimal characters
403 -- with upper case letters. This value is the exclusive or of the
404 -- source checksums of the unit and all its semantically dependent
407 -- The <<attributes>> are a series of two letter codes indicating
408 -- information about the unit:
410 -- DE Dynamic Elaboration. This unit was compiled with the
411 -- dynamic elaboration model, as set by either the -gnatE
412 -- switch or pragma Elaboration_Checks (Dynamic).
414 -- EB Unit has pragma Elaborate_Body
416 -- EE Elaboration entity is present which must be set true when
417 -- the unit is elaborated. The name of the elaboration entity
418 -- is formed from the unit name in the usual way. If EE is
419 -- present, then this boolean must be set True as part of the
420 -- elaboration processing routine generated by the binder.
421 -- Note that EE can be set even if NE is set. This happens
422 -- when the boolean is needed solely for checking for the
423 -- case of access before elaboration.
425 -- GE Unit is a generic declaration, or corresponding body
427 -- IL Unit source uses a style with identifiers in all lower
428 -- IU case (IL) or all upper case (IU). If the standard mixed-
429 -- case usage is detected, or the compiler cannot determine
430 -- the style, then no I parameter will appear.
432 -- IS Initialize_Scalars pragma applies to this unit
434 -- KM Unit source uses a style with keywords in mixed case
435 -- KU (KM) or all upper case (KU). If the standard lower-case
436 -- usage is detected, or the compiler cannot determine the
437 -- style, then no K parameter will appear.
439 -- NE Unit has no elaboration routine. All subprogram bodies
440 -- and specs are in this category. Package bodies and specs
441 -- may or may not have NE set, depending on whether or not
442 -- elaboration code is required. Set if N_Compilation_Unit
443 -- node has flag Has_No_Elaboration_Code set.
445 -- PK Unit is package, rather than a subprogram
447 -- PU Unit has pragma Pure
449 -- PR Unit has pragma Preelaborate
451 -- RA Unit declares a Remote Access to Class-Wide (RACW) type
453 -- RC Unit has pragma Remote_Call_Interface
455 -- RT Unit has pragma Remote_Types
457 -- SP Unit has pragma Shared_Passive.
459 -- SU Unit is a subprogram, rather than a package
461 -- The attributes may appear in any order, separated by spaces.
463 -- ---------------------
464 -- -- W Withed Units --
465 -- ---------------------
467 -- Following each U line, is a series of lines of the form
469 -- W unit-name [source-name lib-name] [E] [EA] [ED]
471 -- One of these lines is present for each unit that is mentioned in
472 -- an explicit with clause by the current unit. The first parameter
473 -- is the unit name in internal format. The second parameter is the
474 -- file name of the file that must be compiled to compile this unit.
475 -- It is usually the file for the body, except for packages
476 -- which have no body; for units that need a body, if the source file
477 -- for the body cannot be found, the file name of the spec is used
478 -- instead. The third parameter is the file name of the library
479 -- information file that contains the results of compiling this unit.
480 -- The optional modifiers are used as follows:
482 -- E pragma Elaborate applies to this unit
484 -- EA pragma Elaborate_All applies to this unit
486 -- ED Elaborate_All_Desirable set for this unit, which means
487 -- that there is no Elaborate_All, but the analysis suggests
488 -- that Program_Error may be raised if the Elaborate_All
489 -- conditions cannot be satisfied. The binder will attempt
490 -- to treat ED as EA if it can.
492 -- The parameter source-name and lib-name are omitted for the case
493 -- of a generic unit compiled with earlier versions of GNAT which
494 -- did not generate object or ali files for generics.
496 -- -----------------------
497 -- -- L Linker_Options --
498 -- -----------------------
500 -- Following the W lines (if any, or the U line if not), are an
501 -- optional series of lines that indicates the usage of the pragma
502 -- Linker_Options in the associated unit. For each appearence of a
503 -- pragma Linker_Options (or Link_With) in the unit, a line is
504 -- present with the form:
508 -- where string is the string from the unit line enclosed in quotes.
509 -- Within the quotes the following can occur:
511 -- c graphic characters in range 20-7E other than " or {
512 -- "" indicating a single " character
513 -- {hh} indicating a character whose code is hex hh (0-9,A-F)
514 -- {00} [ASCII.NUL] is used as a separator character
515 -- to separate multiple arguments of a single
516 -- Linker_Options pragma.
518 -- For further details, see Stringt.Write_String_Table_Entry. Note
519 -- that wide characters in the form {hhhh} cannot be produced, since
520 -- pragma Linker_Option accepts only String, not Wide_String.
522 -- The L lines are required to appear in the same order as the
523 -- corresponding Linker_Options (or Link_With) pragmas appear in
524 -- the source file, so that this order is preserved by the binder
525 -- in constructing the set of linker arguments.
527 ---------------------
528 -- Reference Lines --
529 ---------------------
531 -- The reference lines contain information about references from
532 -- any of the units in the compilation (including, body version
533 -- and version attributes, linker options pragmas and source
536 -- ------------------------------------
537 -- -- E External Version References --
538 -- ------------------------------------
540 -- One of these lines is present for each use of 'Body_Version or
541 -- 'Version in any of the units of the compilation. These are used
542 -- by the linker to determine which version symbols must be output.
543 -- The format is simply:
547 -- where name is the external name, i.e. the unit name with either
548 -- a S or a B for spec or body version referenced (Body_Version
549 -- always references the body, Version references the Spec, except
550 -- in the case of a reference to a subprogram with no separate spec).
551 -- Upper half and wide character codes are encoded using the same
552 -- method as in Namet (Uhh for upper half, Whhhh for wide character,
553 -- where hh are hex digits).
555 -- ---------------------
556 -- -- D Dependencies --
557 -- ---------------------
559 -- The dependency lines indicate the source files on which the compiled
560 -- units depend. This is used by the binder for consistency checking.
561 -- These lines are also referenced by the cross-reference information.
563 -- D source-name time-stamp checksum [subunit-name] line:file-name
565 -- The time-stamp field contains the time stamp of the
566 -- corresponding source file. See types.ads for details on
567 -- time stamp representation.
569 -- The checksum is an 8-hex digit representation of the source
570 -- file checksum, with letters given in lower case.
572 -- The subunit name is present only if the dependency line is for
573 -- a subunit. It contains the fully qualified name of the subunit
574 -- in all lower case letters.
576 -- The line:file-name entry is present only if a Source_Reference
577 -- pragma appeared in the source file identified by source-name.
578 -- In this case, it gives the information from this pragma. Note
579 -- that this allows cross-reference information to be related back
580 -- to the original file. Note: the reason the line number comes
581 -- first is that a leading digit immediately identifies this as
582 -- a Source_Reference entry, rather than a subunit-name.
584 -- A line number of zero for line: in this entry indicates that
585 -- there is more than one source reference pragma. In this case,
586 -- the line numbers in the cross-reference are correct, and refer
587 -- to the original line number, but there is no information that
588 -- allows a reader of the ALI file to determine the exact mapping
589 -- of physical line numbers back to the original source.
591 -- Files with a zero checksum and a non-zero time stamp are in general
592 -- files on which the compilation depends but which are not Ada files
593 -- with further dependencies. This includes preprocessor data files
594 -- and preprocessor definition files.
596 -- Note: blank lines are ignored when the library information is
597 -- read, and separate sections of the file are separated by blank
598 -- lines to ease readability. Blanks between fields are also
601 -- For entries corresponding to files that were not present (and
602 -- thus resulted in error messages), or for files that are not
603 -- part of the dependency set, both the time stamp and checksum
604 -- are set to all zero characters. These dummy entries are ignored
605 -- by the binder in dependency checking, but must be present for
606 -- proper interpretation of the cross-reference data.
608 --------------------------
609 -- Cross-Reference Data --
610 --------------------------
612 -- The cross-reference data follows the dependency lines. See
613 -- the spec of Lib.Xref for details on the format of this data.
615 ----------------------
616 -- Global_Variables --
617 ----------------------
619 -- The table structure defined here stores one entry for each
620 -- Interrupt_State pragma encountered either in the main source or
621 -- in an ancillary with'ed source. Since interrupt state values
622 -- have to be consistent across all units in a partition, we may
623 -- as well detect inconsistencies at compile time when we can.
625 type Interrupt_State_Entry
is record
626 Interrupt_Number
: Pos
;
627 -- Interrupt number value
629 Interrupt_State
: Character;
630 -- Set to r/s/u for Runtime/System/User
632 Pragma_Loc
: Source_Ptr
;
633 -- Location of pragma setting this value in place
636 package Interrupt_States
is new Table
.Table
(
637 Table_Component_Type
=> Interrupt_State_Entry
,
638 Table_Index_Type
=> Nat
,
639 Table_Low_Bound
=> 1,
641 Table_Increment
=> 200,
642 Table_Name
=> "Name_Interrupt_States");
648 procedure Ensure_System_Dependency
;
649 -- This procedure ensures that a dependency is created on system.ads.
650 -- Even if there is no semantic dependency, Targparm has read the
651 -- file to acquire target parameters, so we need a source dependency.
653 procedure Write_ALI
(Object
: Boolean);
654 -- This procedure writes the library information for the current main unit
655 -- The Object parameter is true if an object file is created, and false
658 -- Note: in the case where we are not generating code (-gnatc mode), this
659 -- routine only writes an ALI file if it cannot find an existing up to
660 -- date ALI file. If it *can* find an existing up to date ALI file, then
661 -- it reads this file and sets the Lib.Compilation_Arguments table from
662 -- the A lines in this file.
664 procedure Add_Preprocessing_Dependency
(S
: Source_File_Index
);
665 -- Indicate that there is a dependency to be added on a preprocessing
666 -- data file or on a preprocessing definition file.