Reentrant versions of localtime and gmtime.
Thread-safe programming needs reentrant versions of some of the time functions. The following patch adds localtime_r and gmtime_r a'la Solaris (The Solaris manpage mentiones that these are taken from a POSIX draft, but I don't have a copy to double check). Although it is possible to write localtime and gmtime in terms of the new functions, I decided not to because they are so trival. The Solaris manpage also mentions ctime_r and asctime_r. I'll submit those later tonight. --jtc *** OLD/localtime.c Mon Oct 30 17:00:52 1995 --- localtime.c Mon Oct 30 17:03:48 1995 *************** *** 1026,1031 **** --- 1026,1041 ---- } struct tm * + localtime_r(timep, res) + const time_t * const timep; + struct tm * tm; + { + tzset(); + localsub(timep, 0L, tm); + return tm; + } + + struct tm * localtime(timep) const time_t * const timep; { *************** *** 1072,1077 **** --- 1082,1096 ---- #endif /* State Farm */ } #endif /* defined TM_ZONE */ + } + + struct tm * + gmtime_r(timep, tm) + const time_t * const timep; + struct tm * tm; + { + gmtsub(timep, 0L, tm); + return tm; } struct tm *
Date: Mon, 30 Oct 1995 17:13:53 -0800 From: "J.T. Conklin" <jtc@cygnus.com> struct tm * + localtime_r(timep, res) + const time_t * const timep; + struct tm * tm; + { + tzset(); + localsub(timep, 0L, tm); + return tm; + } That isn't reentrant, since tzset() isn't reentrant. Date: Mon, 30 Oct 1995 17:58:13 -0800 From: "J.T. Conklin" <jtc@cygnus.com> I discovered that most systems declare these as: int asctime_r (const struct tm *, char *, int); int ctime_r (const time_t *, char *, int); while Solaris provides: char *asctime_r (const struct tm *, char *, int); char *ctime_r (const time_t *, char *, int); *AND* char *asctime_r (const struct tm *, char *) char *ctime_r (const time_t *, char *); Sun bug 1185180 claims that the last pair of prototypes appears in Posix.1c draft 10. Solaris 2.5 will be Posix.1c compliant. Sorry, I don't have a copy of Posix.1c -- it hasn't been released to the public yet as far as I know. So I don't know whether the <time.h> interfaces changed again between draft 10 and the official version. I hope Posix.1c casts light on the tzset problem. Also, in Solaris 2.4, <time.h> doesn't declare localtime_r, gmtime_r, etc., unless _REENTRANT is #defined. I don't know why this is. Given the problems with tzset, _REENTRANT, asctime_r, and ctime_r, I suggest getting a copy of the Posix.1c standard before committing to a particular solution here.
participants (2)
-
J.T. Conklin -
Paul Eggert