diff options
Diffstat (limited to 'Build/source/utils/m-tx/mtx-0.60d/libp2c/ChangeLog')
-rw-r--r-- | Build/source/utils/m-tx/mtx-0.60d/libp2c/ChangeLog | 390 |
1 files changed, 390 insertions, 0 deletions
diff --git a/Build/source/utils/m-tx/mtx-0.60d/libp2c/ChangeLog b/Build/source/utils/m-tx/mtx-0.60d/libp2c/ChangeLog new file mode 100644 index 00000000000..17e4bc2dbc5 --- /dev/null +++ b/Build/source/utils/m-tx/mtx-0.60d/libp2c/ChangeLog @@ -0,0 +1,390 @@ +Wed Mar 10 15:03:20 1999 Alexander L. Belikoff <abel@blinky> + + * src/p2c.h: DOUBLE defined as double + +Tue Mar 31 14:57:26 1998 Alexander L. Belikoff <abel@blinky> + + * src/trans.h: + Magic limits (MININT, MAXINT), causing random p2c bugs on the Alpha replaced + with LONG-MIN/MAX from limits.h (integers in Pascal include longs in C, + according to p2c) + +Tue Sep 23 09:14:22 1997 Harvey J. Stein <hjstein@bfr.co.il> + + * src/p2c.proto, src/funcs.c, src/p2c.hdrs: + Adjustments to get pad to work correctly. + +Thu Jul 3 20:41:55 1997 Harvey J. Stein <hjstein@bfr.co.il> + + * src/p2c.proto, src/sys.p2crc, src/trans.h, src/expr.c, src/funcs.c, src/p2c.hdrs: + Bug: Consider: + + procedure foo(x: array[i..j : integer] of varying[k] of char); + ... + var z : array[3..12] of varying[20] of char; + ... + foo(z); + + This should become something like: + + foo(int i, int j, int k, char x[j-i+1][k]) + ... + char z[10][20]; + ... + foo(3, 12, 20, z); + + Instead, it was generating: + + foo(3, 12, 255, z); + + which totally screws up accessing z. + + Analysis: To do this correctly, p2c has to keep dereferencing z as it + processes the hidden arguments i, j, & k so that it's working on an + expression of the appropriate type each time it tries to find either + the size of the string or the start & end points of arrays. Instead, + it just wasn't. + + Fix done: I couldn't figure out exactly how to do the above. But, p2c + calls strmax_func to get the length of the character string. I just + built some additional dereferencing into it so that it gets the + correct length in the above sort of situation. So, changed + strmax_func in expr.c. + + + Bug: Index(x,y) is a common pascal extension for finding a substring + in a string. p2c handles it but it wasn't working right on Dec code. + + Analysis: Some dialects of pascal consider x to be the string and y + to be the substring. Others switch x & y. p2c supports (substring, + string), and Dec pascal supports the opposite. + + Fix: There was some support for switching index argument order, but it + was tightly coupled to HP compiler directive support. I added a + comment switch ("IndexOrder") for additional control of this feature. + Changed func_strpos in funcs.c, and added the switch in trans.h. + +Fri Jun 13 07:22:26 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * src/p2c.hdrs, src/p2c.proto, src/pexpr.c, src/decl.c, src/funcs.c: + Fixed bug in [truncate] handling - I didn't clear the new Meaning + fields (fakenumargs & truncated) at meaning creation time, so they + were sometimes showing up on non-truncated variables. Changed + initmeaning (decl.c). + + Moved hidden number_of_args from beginning to end of arg list. At the + beginning it was interfering with the return hidden variable, + causing return to not be generated. Changes were in p_funcdecl + (decl.c), func_present (funcs.c), and p_funcarglist (pexpr.c). Added + has_truncate function to pexpr.c to check whether an arg list has the + truncated attribute. + + Argument variables were sometimes getting shadowed just because + they're initialized. Problem was in var_reference, which clones + Meanings that have constdefns. But arg variables don't need to be + cloned since the constants are handled by p_funcarglist. So skipped + cloning if Meaning's kind is MK_PARAM or MK_VARPARAM. Previous fix of + always cloning, but only copying meanings if not MK_PARAM seems to + have been incorrect. + +Sun Jun 1 13:15:06 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * src/sys.p2crc, src/trans.h, src/pexpr.c, src/parse.c, src/lex.c, src/p2c.hdrs, src/p2c.proto, src/funcs.c, src/decl.c: + Enhancements: + + Added support for [truncate]. This involved: + 1. Added a truncated flag & a fakenumargs flag to the Meaning + structure. + 2. In p_funcdecl (decl.c), recognizing the [truncate] attribute in + argument lists. If it's seen, the argument's truncated flag is + set, and it and all arguments following are given default values + (assuming that they don't already have default values). The + default values gets p2c to automatically generate the correct + number of arguments for calling the subroutine when some are + left out. + 3. In p_funcdecl (decl.c), if any of the args are tagged with + [truncate], then add a hidden argument to the front of the + argument list - number_of_args. Flag it with fakenumargs. + 4. In p_funcarglist (pexpr.c), when the fcn has a fakenumargs + parameter as the first parameter, (and truncateisdefalt is zero) + compute the index of the last argument supplied, and pass it as + the first argument in the call. + 5. In funcs.c, added func_present to handle the PRESENT(x) calls. + It parses the call, figures out the index of the argument in the + fcns calling sequence (say n), and generates an expression + (number_of_args >= n) to be the output for the PRESENT(x) call. + 6. Added new config parameter - TruncateIsDefault to turn off + PRESENT support (in case one doesn't want the extra arg to be + generated). + + Added support for CONTINUE & BREAK. Just had to basically declare + them as keywords in lex.c, and recognize them in parse.c, and declare + their tokens in trans.h (since p2c already supported them internally). + + Added descriptions for SkipReIncludes, NoOutput, and TruncateIsDefault + to sys.p2crc. + + Bug fixes: + + Bug: args with default values which are also used in local functions + lose their default values. + + Attempted fix: in var_reference (pexpr.c), don't zero out the + constdefn when the arg is cloned if the arg is a parameter. + + Problems: This works, but it sometimes causes bogus declarations of + the old variable to be output. + +Wed May 14 14:27:32 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * src/stuff.c, src/p2c.hdrs, src/p2c.proto, src/out.c, src/expr.c, p2c.patch: + (Updated p2c.patch for last set of changes) + + Bug: p2c was seg faulting when given expressions such as: + there := + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890' + + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890' + + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890' + + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890' + + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890'; + + and: + + there := + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890' + + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890' + + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890' + + '1234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890' + where; + + when there is VARYING [XXX] OF CHAR; + + Analysis: There seem to be all sorts of fixed length character arrays + around the code. It seems that especially long expressions and/or + character strings can cause them to overflow. + + Fix: I upped some of the limits, which at least avoided core dumps on + the code I'm currently processing, but really one needs to go through + and properly keep track of the amount copied into strings so as to + avoid buffer overruns. But, this is a real pain in the ass because of + the shitty semantics of the standard C library. + + Basically, the stack allocated ones that I noticed in expr.c and out.c + I upped to length 2000, and in stuff.c, I upped the size of the + strings in formatbuf[] to 4000. + +Tue May 13 19:27:00 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * src/lex.c, src/p2c.hdrs, src/p2c.proto, src/trans.c, src/Makefile, src/decl.c, src/expr.c, src/funcs.c: + Minor fixes/changes: + -Was calling close in handle_include (lex.c) when it should have + been fclose. + -Modified src/Makefile to allow for using insure on the source code + (to catch memory bugs). + + + Major fixes: + + Bug: p2c occasionally was freeing memory that was already freed. + + Analysis: This seems to occur sometimes on constant string + definitions because mp->val.s might point to the same string as + mp->constdefn->val.s, and forget_ctx tries to free them both. + + Fix: In decl.c, in forget_ctx: After calling free_value(&mp->val) when + mp->kind == MK_CONST, instead of always calling + freexpr(mp->constdefn), only call it when mp->kind != MK_CONST, + or when mp->constdefn->val.s != mp->val.s. + + I also tried to fix this by copying mp->val.s in p_constdecl when + creating mp->constdefn->val.s, but never got this to work + correctly, so I commented it out. + + + Bug: p2c core dumped on: + + readv(data, short_string); + + when data is a packed array[1..7] of char, and + short_string is a packed array[1..3] of char. + + Analysis: What was happening was that in handleread_text (funcs.c), + when the readv argument is an array, it was checking + ord_range(ex->val.type->indextype, ...), and ex was NULL. + + Fix: It seems like this was a typo, so I change it to check + var->... instead of ex->... + + + Bug: p2c core dumped on: + + WRITELN('%SQLMOD ', msg); + + when msg was of type varying[x] of char. + + Analysis: Some part of the code was converting the above string to + "%%SQLMOD %s". It seems that cleansprintf (from expr.c) didn't + realize that % is escaped by %%, so it thought that there should + be 2 arguments (maybe 3?) for the above printf string instead of + one, and thus was going off of the end of the args list (i.e. - + ex->args[j] out of range). + + Fix: Inside the for loop, if the next character is also a %, then + skip arg processing, and skip the next character (just copy it + into fmtbuf). + + + Bug: p2c core dumps on: + + readv(c, i); + + when c is a char and i is an integer. + + Analysis: In funcs.c, in handleread_text, var is never getting + initialized - it's getting passed in as NULL, and it only gets + set if isstrread is true. Since c is a char, fex->val.type->kind + is TK_CHAR, not TK_STRING, so isstrread is false, so var fails to + get initialized. + + Fix: After the case that var is set when isstrread is true, if it's + still NULL, we set it to p_expr(NULL). This prevents the core + dump, but produces incorrect code! It outputs: + + fscanf(c, "%ld", &i); + + when it should more likely be something along the lines of + sscanf(&c, ...) (but safer). + +Mon May 5 15:23:04 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * src/out.c, src/p2c.hdrs, src/p2c.proto, src/trans.h, src/decl.c, src/lex.c, p2c.patch: + New features (out.c, trans.h): + New parameter NoOutput - setting it to 1 stops p2c from generating + output. Useful when trying to simulate Dec style modules via + include files - just convert INHERIT to %include, and put + {*NoOutput=1} at the top of each include file, and p2c does the + right thing. This was done by not pushing characters into the + output stream when nooutput is 1. The real code changes occur in + out.c, in putc_outf, and puts_outf. Accumulating the output, but + turning off writes (i.e. - wiring it into flush_outfilebuf) failed + because of timing differences btw when comment directives are + processed and when output occurs. + + Bug fixes: + + Bug: Some variable declarations weren't getting output. Most + noticable were the automatically generated Return variables. + + Analysis: When using the {INTF-ONLY} directive in include files + then the fwd declaration of fcns was being interpreted as the fcn + definition, and so the variables generated when the declaration was + seen were marked as "wasdeclared". Then, when the code for the + function was being generated, these variables weren't getting + output because "wasdeclared" was 1. This was most visible from the + automatically generated Return variable, but occasionally happened + to other variables. + + Fix: in decl.c:activatemeaning - added resetting of "wasdeclared" + if we can identify that the variable is the autogenerated return + variable. Detected them by snarfing code from isretvar. But this + was a lousy fix because as mentioned above, there are other + variables which are occasionally affected. And, it's no longer + needed because of the new NoOutput feature, so I commented out the + change, but left the code there as a reminder of the problem. + + + Bug: Comments weren't getting output in the correct places. This + is a serious problem when using {EMBED comments to insert C code + into the output stream. Sometimes code from the {EMBED comment + given in an include file showed up in the includer. Sometimes the + code didn't get emitted at all because of mistiming against + {nooutput} directives. + + Analysis: p2c builds up a comment string structure as it parses + the code, and hangs it off of the code's internal representation. + Presumably this is so that when doing code transformations, the + comments are output in the correct places. However, this seems to + sometimes fail when using %include files. It seems that the + comments weren't getting processed & output until the current block + was finished. Included files don't start a new block because one + could conceivably include random snippets of code anywhere you want + (inside of RECORD declarations, for example). Thus, sometimes + comments from inside the include file don't get processed & output + until well after the include file itself is finished being + processed. + + Fix: Added the call flushcomments(NULL, -1, -1) to both push_input + and pop_input (in lex.c). This seems to have fixed it with no + noticable side effects. + +Tue Apr 29 18:36:51 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * src/p2c.hdrs, src/p2c.proto, src/lex.c: + Added recognition of AND_THEN and OR_ELSE. + + * src/trans.h, src/lex.c, src/p2c.hdrs, src/p2c.proto: + Added tracking of files that have been included so that they can + optionally be ignored the second time they're seen. This is needed + because a module (say X) can inherit from two other modules (say Y and + Z), both of whom inherit from a fourth module (say Z). Converting the + inherits to includes thus gets Z included twice. If it's not ignored, + it causes p2c to start renaming procedures so as to avoid name + clashes. + + Added an rc file parameter (skipreincludes) for controling the above + action. + + * src/decl.c, src/p2c.hdrs, src/p2c.proto, src/parse.c: + Various adjustments & bug fixes suggested by Dave Gillespie: + + decl.c: Call format_s so that we work with a copy of the string + when removing dollar signs - a little more robust. + + Missing ! on strcicmp caused bogus externs to be emitted on + some variable declarations. + + parse.c: Sprinkle afew p_attributes around p_block so that p2c + properly parses attributes in front of declaration blocks. + In particular, this fixes the problem where p2c was choking + on [hidden] followed by TYPE, which appears in analytic.pas. + +Wed Apr 23 12:17:08 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * src/p2c.proto, src/decl.c, src/p2c.hdrs: + Modified addfield so that it removes dollar signs from record field + identifiers. This is more consistent with the handling of other + identifiers, and also needed by much of the Pascal code. + +Mon Apr 21 10:41:53 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * p2c.patch: Submitted new patch to Dave G. + + * src/p2c.proto, src/lex.c, src/p2c.cat, src/p2c.hdrs, src/Makefile: + Changed VALUE from a keyword to a possible keyword when working with + Dec Pascal code. Original code set it to be a keyword for Dec, and a + possible keyword otherwise. However, if it's a keyword, then p2c + doesn't accept it the name of a slot in a record. Since Dec accepts + this usage, I figured that VALUE should always be a possible keyword + instead of a keyword. + +Thu Apr 17 11:33:22 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * p2c.patch: patches generated & mailed to Dave Gillespie + +Wed Apr 16 16:26:07 1997 Harvey J. Stein <hjstein@blinky.bfr.co.il> + + * src/p2c.hdrs, src/p2c.proto, .cvsignore, examples/.cvsignore, src/.cvsignore, src/decl.c, src/p2c.cat: + Parser didn't expect a ( after the OTHERWISE keyword. Modified + decl.c to expect it. + + * src/INSTALL, src/citmods.c, src/comment.c, src/decl.c, src/dir.c, src/expr.c, src/funcs.c, src/hpmods.c, src/lex.c, src/loc.p2clib.c, src/loc.p2crc, src/makeproto.c, src/out.c, src/p2c.cat, src/p2c.h, src/p2c.hdrs, src/p2c.proto, src/p2cc.cat, src/p2cc.man, src/p2cc.perl, src/p2clib.c, src/parse.c, src/pexpr.c, src/string.pas, src/system.imp, src/system.m2, src/trans.h, src/turbo.imp: + Import of p2c 1.21alpha2 + + * src/INSTALL, src/citmods.c, src/comment.c, src/decl.c, src/dir.c, src/expr.c, src/funcs.c, src/hpmods.c, src/lex.c, src/loc.p2clib.c, src/loc.p2crc, src/makeproto.c, src/out.c, src/p2c.cat, src/p2c.h, src/p2c.hdrs, src/p2c.proto, src/p2cc.cat, src/p2cc.man, src/p2cc.perl, src/p2clib.c, src/parse.c, src/pexpr.c, src/string.pas, src/system.imp, src/system.m2, src/trans.h, src/turbo.imp: + Initial revision + + * HP/import/asm.imp, HP/import/fs.imp, HP/import/general_0.imp, HP/import/general_1.imp, HP/import/general_2.imp, HP/import/general_3.imp, HP/import/general_4.imp, HP/import/hpib_0.imp, HP/import/hpib_1.imp, HP/import/hpib_2.imp, HP/import/hpib_3.imp, HP/import/iodecl.imp, HP/import/misc.imp, HP/import/rnd.imp, HP/import/sysdevs.imp, HP/import/sysglobals.imp, HP/include/asm.h, HP/include/fs.h, HP/include/general_0.h, HP/include/general_1.h, HP/include/general_2.h, HP/include/general_3.h, HP/include/general_4.h, HP/include/hpib_0.h, HP/include/hpib_1.h, HP/include/hpib_2.h, HP/include/hpib_3.h, HP/include/iodecl.h, HP/include/misc.h, HP/include/rnd.h, HP/include/sysdevs.h, HP/include/sysglobals.h, Makefile, README, examples/Makefile, examples/README, examples/basic.doc, examples/basic.p, examples/c/basic.c, examples/c/cref.c, examples/c/e.c, examples/c/fact.c, examples/c/self.c, examples/cref.p, examples/e.p, examples/fact.p, examples/p2crc, examples/self.p, ftp.README, src/COPYING, src/HISTORY, src/Makefile, src/NOTES, src/README, src/p2c.man, src/stuff.c, src/sys.p2crc, src/trans.c: + Import of p2c 1.21alpha2 + + * HP/import/asm.imp, HP/import/fs.imp, HP/import/general_0.imp, HP/import/general_1.imp, HP/import/general_2.imp, HP/import/general_3.imp, HP/import/general_4.imp, HP/import/hpib_0.imp, HP/import/hpib_1.imp, HP/import/hpib_2.imp, HP/import/hpib_3.imp, HP/import/iodecl.imp, HP/import/misc.imp, HP/import/rnd.imp, HP/import/sysdevs.imp, HP/import/sysglobals.imp, HP/include/asm.h, HP/include/fs.h, HP/include/general_0.h, HP/include/general_1.h, HP/include/general_2.h, HP/include/general_3.h, HP/include/general_4.h, HP/include/hpib_0.h, HP/include/hpib_1.h, HP/include/hpib_2.h, HP/include/hpib_3.h, HP/include/iodecl.h, HP/include/misc.h, HP/include/rnd.h, HP/include/sysdevs.h, HP/include/sysglobals.h, Makefile, README, examples/Makefile, examples/README, examples/basic.doc, examples/basic.p, examples/c/basic.c, examples/c/cref.c, examples/c/e.c, examples/c/fact.c, examples/c/self.c, examples/cref.p, examples/e.p, examples/fact.p, examples/p2crc, examples/self.p, ftp.README, src/COPYING, src/HISTORY, src/Makefile, src/NOTES, src/README, src/p2c.man, src/stuff.c, src/sys.p2crc, src/trans.c: + Initial revision + |