Some olders news from the gfortran development
- Steve Kargl added shape-conformance checking for intrinsic operations, see [gccbug:19754] (2005-03-05)
- Tobias Schlüter fixed a bug which made the compiler not dereference pointers correctly under some unusual circumstances, see [gccbug:19673] (2005-03-04)
- Tobias Schlüter ported a corrected constant-folder for the UBOUND and LBOUND intrinsics, see [gccbug:19479] (2005-03-01)
- Steve Kargl fixed an off-by-one issue with the IARGC intrinsic (2005-02-28)
- François-Xavier Coudert fixed printing of zero when P formats are in use, see [gccbug:20101] He also found a wrong condition which made REAL and DOUBLE PRECISION print the same exponent character (2005-02-27)
- Steve Kargl implemented Fortran 2003's rules for integer boz expressions, see [gccbug:20058] (2005-02-27)
- Peter O'Gorman fixed build issues on Mac OS X, see [gccbug:17748] (2005-02-25)
- Bud Davis fixed a problem where output would act strangely with hollerith constants, see [gccbug:20086] (2005-02-21)
- Eric Botcazou fixed problems with infinities on SPARC, see [gccbug:19302] (2005-02-21)
- Steve Kargl fixed issues with the ACHAR, INT, SELECTED_INT_KIND, CEILING, FLOOR and DATE_AND_TIME intrinsics (2005-02-19)
- Tobias Schlüter fixed a bug with substrings of constant strings (2005-02-08)
- Tobias Schlüter ported some enhancements which give better error reporting with conflicting attributes (2005-02-07)
- Paul Thomas fixed issues with infinities and NANs? in I/O, see [gccbug:19363] (2005-02-01)
- Richard Henderson changed gfortran to use math routines which don't set the C-only errno, therefore speeding up generated code (2005-01-30)
- Bud Davis fixed a problem where DELIM inquiries would return garbage, see [gccbug:19647] (2005-01-30)
- Thomas Koenig added checking that EOR is only used in ADVANCE='YES' I/O, see [gccbug:19595] (2005-01-29)
- Steve Kargl added conformance checks to LOGICAL assignments, see [gccbug:19589] (2005-01-29)
- Steven Bosscher updated the handling of compiler-generated labels (2005-01-27)
- Tobias Schlüter fixed a case where the library would free unallocated memory, see [gccbug:19524] (2005-02-25)
- Steve Kargl and Paul Brook made the compiler accept COMPLEX literal constants with blanks in them, see [gccbug:17941]
- James Morrison fixed alignment issues with the TRANSPOSE intrinsic and COMPLEX arrays, see [gccbug:19294] (2005-01-23)
- Thomas Koenig made the library not write to read-only files, even if requested, see [gccbug:19451] (2005-01-22)
- Steve Kargl fixed issues with the ACHAR, INT, SELECTED_INT_KIND, CEILING, FLOOR and DATE_AND_TIME intrinsics (2005-02-19)
- Tobias Schlüter fixed a bug with substrings of constant strings (2005-02-08)
- Tobias Schlüter ported some enhancements which give better error reporting with conflicting attributes (2005-02-07)
- Paul Thomas fixed issues with infinities and NANs? in I/O, see [gccbug:19363] (2005-02-01)
- Richard Henderson changed gfortran to use math routines which don't set the C-only errno, therefore speeding up generated code (2005-01-30)
- Bud Davis fixed a problem where DELIM inquiries would return garbage, see [gccbug:19647] (2005-01-30)
- Thomas Koenig added checking that EOR is only used in ADVANCE='YES' I/O, see [gccbug:19595] (2005-01-29)
- Steve Kargl added conformance checks to LOGICAL assignments, see [gccbug:19589] (2005-01-29)
- Steven Bosscher updated the handling of compiler-generated labels (2005-01-27)
- Tobias Schlüter fixed a case where the library would free unallocated memory, see [gccbug:19524] (2005-02-25)
- Steve Kargl and Paul Brook made the compiler accept COMPLEX literal constants with blanks in them, see [gccbug:17941]
- James Morrison fixed alignment issues with the TRANSPOSE intrinsic and COMPLEX arrays, see [gccbug:19294] (2005-01-23)
- Thomas Koenig made the library not write to read-only files, even if requested, see [gccbug:19451] (2005-01-22)
- David Edelsohn preconnected unit 0 to stderr, see [gccbug:19052] (2005-01-22)
- Thomas Koenig added an error for trying to open an existing file with STATUS="NEW", see [gccbug:18982]
- Bud Davis implemented the PAD and POSITION inquiries, see [gccbug:19313] and [gccbug:19314] (2005-01-22)
- David Edelsohn committed a patch which preconnects unit 0 to stderr, see [gccbug:19052] (2005-01-22)
- Thomas Koenig fixed a bug where opening a file with STATUS='NEW' would not issue an error, see [gccbug:18982] (2005-01-22)
- Steve Kargl fixed some bugs in intrinsic procedures (2005-01-22)
- Paul Brook the typing of implicitly typed contained functions (2005-01-22)
- Tobias Schlüter fixed a type mismatch between the library and the compiler which made some I/O statements fail on 64-bit platforms. See [gccbug:19194] (2005-01-22)
- Several patches have been made which internal type mismatches. We don't know of any visible effect of these. (2005-01-15)
- Bud Davis fixed a bug where to long input lines would break with A formats, see [gccbug:18398] (2005-01-07)
- Tobias Schlüter made some bit intrinsics work like they did in g77 for cases not specified by the standard. He also fixed the MODULO intrinsic, which was broken in the case of a unit divisor, see [gccbug:19032] (2004-12-27)
- Andrew Pinski relaxed restrictions that were made to the evaluation of integral powers (2004-12-27)
- Bud Davis fixed a bug in the formatted output of complex numbers, see [gccbug:19071] (2004-12-23)
- Steve Kargl fixed a typo in the gfortran info page (2004-12-23)
- Aaron W. La Framboise fixed a header conflict which prevented building the compiler library on MingW? for the last few days, see [gccbug:19074] (2004-12-19)
- Tobias Schlüter fixed an omission in the matching of the NULLIFY statements which lead to weird error messages, see [gccbug:18993] (2004-12-15)
- Richard Henderson updated parts of the compiler to use gcc's common symbol handling facilities (2004-12-14)
- Steve Ellcey fixed an issue where a misaligned memory access in the library would cause segmentation faults (2004-12-14)
- Steve Kargl implemented some of the bit intrinsics for single-byte and two-byte integers, see [gccbug:18966] (2004-12-14)
- David Edelsohn fixed a minor typo in the library (2004-12-14)
- Tobias Schlüter fixed problems where the parsing of multiple common blocks in a single COMMON statement in free-form source code would fail, see [gccbug:18869] (2004-12-12)
- Steve Kargl fixed problems with the bit intrinsics and non-default INTEGER kinds, see [gccbug:16581] (2004-12-12)
- Steve Kargl and Paul Brook implemented the common extension of allowing REAL DO-loop iterators, see [gccbug:16222] (2004-12-12)
- Richard Henderson committed a patch which creates a defined external interface to libgfortran, the runtime library of gfortran, which cleanly prevents namespace pollution (2004-12-12)
- Tobias Schlüter fixed some problems where gfortran was too permissive when the old time Fortran ASSIGN statement was used (2004-12-08)
- Tobias Schlüter made a change to give better loci for errors in END statements (2004-12-08)
- Steve Ellcey committed a patch which made checks while opening files a bit more permissive, see [gccbug:18653]. He also fixed issues with cross-builds. (2004-12-07)
- Eric Botcazou fixed some longstanding issues with function prototypes that shows up on non C99 systems (2004-12-07)
- Richard Henderson did some code clean up in the library, and fixed a problem with the auto-generation of some of the library files (2004-12-06)
- Tobias Schlüter fixed unformatted I/O of COMPLEX numbers, see [gccbug:18710]. This together with a previous fix to the INQUIRE statement from Bud Davis fixes NIST test FM910?.FOR (2004-12-02)
- Steve Kargl and Paul Brook implemented a couple more of g77's nonstandard intrinsics, FNUM, STAT, FSTAT, and FLUSH (2004-12-02)
- Bud fixed the last bug in NIST test FM413?.FOR (2004-11-28)
- Bud Davis fixed a bug in the ENDFILE statment, see [gccbug:18364] (2004-11-27)
- Paul Brook fixed the layouting of selfreferential types, see [gccbug:13010] (2004-11-16)
- Victor Leikehman submitted some work to speed up the MATMUL intrinsic (2004-11-14)
- Paul Brook fixed a problem where memory was deallocated in the wrong order leading to segfaults, see [gccbug:18375] (2004-11-10)
- Paul Brook enhanced the routines for reading REAL input. This should cure precision problems, such as those in [gccbug:18218] (2004-11-10)
- Tobias Schlüter fixed a bug where a common typo in PRINT statements would cause a segfault, see [gccbug:18023] (2004-11-06)
- Tobias Schlüter fixed a bug that his previous patch made reappear, see [gccbug:15164] (2004-11-05)
- Tobias Schlüter fixed a few bugs with generic interfaces in modules, see [gccbug:17535], [gccbug:17583], [gccbug:17713] (2004-11-03)
- Janne Blomqvist added code that determines if a user-referenced procedure is an intrinsic according to the language standard the user selected, see [gccbug:17590] (2004-11-02)
- Tobias Schlüter fixed a bug in the debug info which made code blocks appear in the wrong places when disassembling (2004-10-30)
- Canqun Yang and Paul Brook fixed RAND and IRAND to have an optional argument (2004-10-30)
- Scott Robert Ladd patched gfortran to allow the full range of twos-complement integers, instead of only those defined by the integer model of the standard, see [gccbug:13490], and [gccbug:17912] (2004-10-29)
- Andrew Pinski fixed a bug in the -fmax-identifier-length= compiler option, see [gccbug:17901] (2004-10-08)
- Tobias Schlüter fixed a bug in the error reporting for user defined operators which caused segfaults on some platforms, see [gccbug:17676] (2004-10-08)
- Paul Brook plugged a memory leak when writing to internal files (2004-10-07)
- Paul Brook fixed an off-by-one error in the VERIFY intrinsic, which previously ignored the last character in a string (2004-10-07)
- Paul Brook fixed a bug where use-associated variables were erroneously initialized, see [gccbug:17678] (2004-10-07)
- Tobias Schlüter checked in a patch which allows kind parameter suffixes to boz literal constants (which are allowed everywhere as an extension already). This allows entering integers in binary/octal/hexadecimal even if they are bigger than default integer (2004-10-07)
- Tobias Schlüter changed constant folding of the CABS intrinsic to use a numerically safe algorithm (2004-10-06)
- Paul Brook fixed an issue which affected the skipping of records in list-directed reads, see [gccbug:17709] (2004-10-06)
- Paul Brook enhanced the code generation for DO loops. Much better code should result from this (2004-10-06)
- Paul Brook and Bud Davis fixed two bugs with formatted I/O, one with printing negative zero, the other with repeat counts not being set for certain formats see [gccbug:16434] and [gccbug:17706] (2004-10-04)
- Tobias Schlüter fixed a series of bugs: the PACK intrinsic didn't work with a scalar mask, also it didn't allocate memory for its return value which is necessary in some cases, see [gccbug:17283], he implemented the MVBITS intrinsic, which was missing, see [gccbug:17631], he fixed a problem where a GOTO to an END DO would go out of the loop, see [gccbug:17708], finally he implemented the SYSTEM intrinsic, which allows executing shell commands from a Fortran program, see [gccbug:17776] (2004-10-03)
- Erik Schnetter fixed a bug in the scanner when dealing with preprocessed source files, see [gccbug:17734] (2004-10-03)
- Paul Brook and Tobias Schlüter fixed a bug where the character length of a component was not determined correctly, leading to internal errors, see [gccbug:17612] (2004-10-03)
- Tobias Schlüter, together with Patrick Pelissier, reworked some of the compiler's internal arithmetic to give more precise results in less time (2004-10-03)
- Feng Wang added constant folding to the implementation of RRSPACING (2004-10-03)
- Tobias Schlüter found that the fixing of two old bugs seems to have gone unnoticed. New testcases were added to the testsuite to make sure that they remain fixed, see [gccbug:15164] and [gccbug:16938] (2004-09-27)
- Tobias Schlüter fixed an issue where gfortran relied on the powf function, which is not available on all systems, see [gccbug:16137] (2004-09-26)
- Tobias Schlüter fixed a problem with generic interfaces where the code generation passes would look at the wrong symbol and then issue an internal error under certain circumstances, see [gccbug:17615] (2004-09-24)
- Tobias Schlüter fixed a problem where using the ETIME intrinsic yielded a link error when it was declared EXTERNAL. Code that uses this appears in LAPACK (2004-09-24)
- Steven Kargl fixed a problem where incompatible declarations of alloca broke the build on some platforms (2004-09-21)
- Bud Davis fixed a problem with reading namelists, see [gccbug:17286] (2004-09-21)
- Tobias Schlüter fixed a problem where constant folding of calls to RESHAPE with the optional ORDER argument returned an array of the wrong shape, see [gccbug:15957] (2004-09-20)
- Paul Brook has fixed a compatibility issue where older versions of gcc were not able to compile gfortran. This is mostly an issue for users of cross-compilers where a bootstrapped build would add additional complication (2004-09-17)
- Victor Leikehman and Paul Brook have added code for printing array components of derived types, see [gccbug:15364] (2004-09-16)
- Aaron W. La Framboise has submitted a set of patches for running gfortran on MingW?, which allows using gfortran on Windows systems without having to install cygwin (2004-09-16)
- Steven Kargl has implemented the GETCWD intrinsic which brings gfortran one step closer to supporting g77's language extensions (2004-09-15)
- Tobias Schlüter and Paul Brook fixed two bugs with PRIVATE symbols from modules, see [gccbug:16485] and [gccbug:16609] (2004-09-15)
- Bud Davis fixed a bug where trailing spaces confused the routines for list-directed reading, see [gccbug:17090] (2004-09-14)