v0.3 (2007-05-01)

I have tagged and tarballed a 0.3 release of Sparse, now available from http://ftp.be.debian.org/pub/software/devel/sparse/dist/ with sha1sum 1d868b29234176abd5f3f5463aad1f11d5268dc2.

Note that the Sparse Git repository has moved to:

The old repository location will continue to work for now, but please update any references you have to the old location.

Thanks to Christopher Li for contributing heavily to this release, including several notable new features. See the full changelog for details.

In addition to numerous bug fixes, cleanups, and new test cases, this release includes several new visible features:

  • A Sparse-based implementation of ctags, thanks to Christopher Li. Now you can have a ctags that ‘actually parses C’.
  • Sparse now correctly supports annotations on inline functions.
  • The new ‘-ventry’ option will make Sparse dump its internal linearized bytecode format.
  • Sparse now parses the ‘__regparm__’ attribute just like ‘regparm’, fixing a warning with pthreads from glibc 2.5.
  • Sparse now interprets the integer size attributes ‘QI’, ‘SI’, ‘HI’, and ‘DI’, and their double-underscore variants.
  • Sparse now supports attributes on labels, sometimes used to specify ‘unused’ labels.
  • Sparse now handles structure attributes between the structure keyword and the name, fixing many parse errors.
  • Sparse now supports more than one command-line include file.
  • The pkg-config file gets installed to its correct location under ‘lib’, rather than under ‘share’.

Notable internal changes:

  • Fully table-driven attribute parsing. This simplifies the C parser and makes it far easier to add new attributes to Sparse.
  • Many internal cleanups, more information preserved for backends, more useful formats for that information, and lower memory usage.
  • Fixed Sparse warnings about itself.

Full changelog:

Christopher Li (24):
  • Sparse-based Ctags implementation
  • Change the symbol access list to a pseudo list
  • Add instruction to pseudo user tracking.
  • Update usage chain for dead instructions
  • Update usage chain for dead branch instruction.
  • Allow more than one command line include file.
  • Enhance debug information.
  • Another attempt to fix the attribute parsing.
  • Marking anonymous string.
  • Bug fix in pointer modifier inheritance at function degeneration.
  • Handle structure attributes between the structure keyword and the name
  • Fix the segfault when initializer has unknown symbol
  • Fix double semicolon in struct declaration
  • Make the ptrlist using the sparse allocator.
  • Fix core dump on anonymous symbol.
  • Fix a bug that match_idents forget to end with NULL
  • Adding debug option for showing the linearized instruction.
  • Fix core dump on huge switch
  • Introduce expression_error
  • Disable liveness “dead” instruction by default.
  • Add annotation for inline function call.
  • Introduce keyword driven attribute parsing
  • Fix the annotated inline call position
  • handle label attributes
Christopher Li and Josh Triplett (4):
  • Introduce top level parsing for asm parsing.
  • Introducing statement keywords
  • Free up some special bits in modifiers.
  • Moving statement parsing into smaller functions.
James Westby (2):
  • Fix mistaken comparison that becomes a no-op.
  • Update the information in README about using the library.
Josh Triplett (30):
  • Add ctags to .gitignore
  • Add a return in the last case of a switch; redundant but less error-prone.
  • Coding style fix: in a pointer type, * goes with the name, not the type.
  • Add missing #include “allocate.h” in linearize.h for DECLARE_ALLOCATOR.
  • Add test case for function pointer modifier inheritance
  • Add test case for structure attribute placement.
  • Add test case for double semicolon in structure declaration.
  • Coding style fix: use parentheses with sizeof
  • Move pkg-config file to lib, rather than share
  • Add static to declarations in test cases, to remove unrelated warnings.
  • Fix typo in symbol.h: s/keywrods/keywords/
  • Fix typos in comments
  • Use GCC format and sentinel attributes on appropriate functions
  • Fix two potential NULL pointer dereferences in dissect.c
  • Avoid returning an uninitialized pointer from dup_list of an empty list
  • Remove stray space from expand_compare in expand.c
  • Prevent potential NULL pointer dereference in expand_compare
  • Add test case for basic address_space annotations.
  • Use noreturn on die() and error_die()
  • Parse and ignore the __regparm__ attribute, just like regparm.
  • Fix comment to reference #weak_define rather than #ifndef, matching code
  • Teach cgcc about -Wtransparent-union and -Wno-transparent-union
  • Declare die_if_error extern in lib.h
  • Remove unused variable “include” from lib.c
  • Declare do_error static
  • Declare gcc_patchlevel extern in lib.h
  • compile-i386.c: Declare regs_in_use static
  • simplify.c: Declare delete_pseudo_user_list_entry static
  • linearize: DECLARE_ALLOCATOR for asm_constraint and asm_rules
  • Fix most -Wshadow warnings in Sparse.
Oleg Nesterov (3):
  • dissect: cleanup report_implicit()
  • dissect: fix multidimensional array initializer
  • dissect: simplify lookup_member()

– Josh Triplett