Type confusion and strict aliasing in gtfringetable

[Michael Wood-Vasey]

Paul says
"""
off_t is probably defined as "long int" instead of "long long int", in which case the cast is inappropriate.

The quickest workaround would be to turn of -Werr in the Makefile.
"""

but this looks general so should be fixed.

Joel reported to this to the ps-ipp-dev list last week.

"""
/home/giga/temp/PS1/src/ipp/Ohana/src/misc/src/gtfringetable.c: In function ‘main’:
/home/giga/temp/PS1/src/ipp/Ohana/src/misc/src/gtfringetable.c:38: warning: dereferencing type-punned pointer will break strict-aliasing rules
/home/giga/temp/PS1/src/ipp/Ohana/src/misc/src/gtfringetable.c:39: warning: dereferencing type-punned pointer will break strict-aliasing rules
/home/giga/temp/PS1/src/ipp/Ohana/src/misc/src/gtfringetable.c:40: warning: dereferencing type-punned pointer will break strict-aliasing rules
/home/giga/temp/PS1/src/ipp/Ohana/src/misc/src/gtfringetable.c:41: warning: dereferencing type-punned pointer will break strict-aliasing rules
"""

[eugene]

This issue is caused by the difficulty of printing/scanning data of off_t type. Since the off_t type can vary in size between systems, a single print format cannot be used. This specific issue was driven by the inappropriate use of casting to get around this problem -- the casting sometimes violated optimization rules.

This issue has been addressed by reworking the code to use an implementation-dependent print format code 'OFF_T_FMT', which depends on the state of the LARGEFILE configuration. This macro is sometimes set of %ld and sometimes to %jd (appropriate for intmax_t).

fix is on the trunk in r28241