home *** CD-ROM | disk | FTP | other *** search
-
-
-
- - 1 -
-
-
-
- Base Development 7.2.1 Release Notes
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
- - 2 -
-
-
-
- DDDDooooccccuuuummmmeeeennnntttt NNNNuuuummmmbbbbeeeerrrr 000000008888----1111777788882222----000033330000
-
-
- 5. _B_u_g__F_i_x_e_s
-
- This chapter briefly describes the bugs that
- have been fixed in 7.2.1 compiler_dev since
- release 7.2. Some of the headings are followed
- by a Silicon Graphics incident report number.
-
- 5.1 _l_d
-
- +o ld has been fixed to pad .MIPS.stubs code
- with a nop to avoid end of page bugs in
- MIPS microprocessors. (Bug #454860).
-
- +o With the 7.2 MR release of the compilers on
- IRIX 6.2, ld would fail under certain
- circumstances with the message:
-
- ld32: FATAL 2: Internal: at ../../ld/layout_util.c
- undefined symbols merged into symbol table
-
- This has been fixed (Bug #537731).
-
- +o With the 7.2 MR release of the compilers on
- IRIX 6.2, ld would fail under certain
- circumstances with the message:
-
- ld32: FATAL 2: Internal: at ../../ld/write.c
- there are sections which overlap or are out of order
-
- This has been fixed (Bug #528270).
-
- +o The ld option -default_delay_load can
- occasionally cause problems. In certain
- cases, when an application (MAIN)
- references a dso (libONE.so) which contains
- a function which references a global
- variable (FOO) defined in another dso
- (libTWO.so) which has been built with
- -default_delay_load, the application may
- show the following error:
-
- 1374:MAIN :rld: Fatal Error: unresolvable symbol in libONE.so: FOO
-
- In such cases, ld is supposed to override
- the delay_load marking for libTWO.so and
- mark it as immediate load. However,
- sometimes it fails to do this. The
- workaround is build libTWO.so without using
-
-
-
-
-
-
-
-
-
-
-
- - 3 -
-
-
-
- -default_delay_load. This has been closed.
- (Bug #494294)
-
-
- +o ld options -delay_load and
- -default_delay_load can slow down run-time
- performance of certain applications. This
- is because rld will try to re-resolve
- undefined symbols in the main executable
- and all currently loaded dsos. This has
- been fixed. (Bug #506076)
-
-
- +o Unable to locate -Xlocal symbol when
- compiling non-shared. This has been fixed
- (Bug #328118).
-
- +o ld generates bogus error msg with a strange
- so_locations file. This has been fixed
- (Bug #338569).
-
- +o EndOfPage warning should state the name of
- module. This has been fixed (Bug #346433).
-
- +o When -D linker option specified, should
- (16K) page align. This has been fixed (Bug
- #364037).
-
- +o multigot linker -O3 -mp gets Bus Error at
- routine call. This has been fixed (Bug
- #381418).
-
- +o Fails to zero __Argv in a.out, causing
- ssrun -pcsamp crashes. This has been fixed
- (Bug #389942).
-
- +o Warn about missing Fortran trailing
- underscores. This has been fixed (Bug
- #433633).
-
- +o ld & delay_load cores dump with C++. This
- has been fixed (Bug #459057).
-
- +o -delay_load not working on 7.1.1 -n32 -O3.
- This has been fixed (Bug #481637).
-
- +o Mixing DATA statements and -Xlocal on a
- COMMON leads to ld error. This has been
- fixed (Bug #494002).
-
-
-
-
-
-
-
-
-
-
-
-
-
- - 4 -
-
-
-
- +o Delay load failures do not inform/educate
- users. This has been fixed (Bug #507824).
-
- +o Only warn about message numbers legal for
- linker but not yet used. This has been
- fixed (Bug #520795).
-
- +o -multigot created sections have symbols
- with null symtab name. This has been fixed
- (Bug #521276).
-
- +o Bad compact relocation records generated by
- o32 7.2 ld. This has been fixed (Bug
- #522040).
-
- +o -abi causes new multigot linker to abort.
- This has been fixed (Bug #525061).
-
- +o -32 link with use defined hiddens causes
- the linker to fail. This has been fixed
- (Bug #526292).
-
- +o corded executable bad due to linker got
- page tiling. This has been fixed (Bug
- #526689).
-
- +o Internal: at ../../ld/multigot.c in
- mgot_update_scommons().... This has been
- fixed (Bug #527541).
-
- +o Target address outside of segment cause
- linker to fail. This has been fixed (Bug
- #527611).
-
- +o -32 bogus lit section entsizes confused the
- linker. This has been fixed (Bug #527614).
-
- +o Seg fault on non-SGI (gnu) compiled xanim
- object. This has been fixed (Bug #551389).
-
- +o Compile generates bad a.out This has been
- fixed (Bug #526618).
-
- +o 7.2 Linker Seg Faults This has been fixed
- (Bug #538341).
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-