Re: getenv() and i370-ibm-openedition

[Thomas Dickey]

On Wed, Jul 23, 2014 at 04:26:55PM -0600, Paul Gilmartin wrote:
> Hello, bug-ncurses,
> 
> In: 
> http://pubs.opengroup.org/onlinepubs/9699919799/functions/getenv.html#tag_16_194_03
> one may read:
>     ...
> |char *getenv(const char */name/);|
>     ...
> ^[CX <javascript:open_code('CX')>] [Option Start] The returned
> string pointer might be invalidated or [Option End]  the string
> content might be overwritten by a subsequent call to /getenv/(),
> ^[CX <javascript:open_code('CX')>] [Option Start] /setenv/() 
> <http://pubs.opengroup.org/onlinepubs/9699919799/functions/setenv.html>,
> /unsetenv/() 
> <http://pubs.opengroup.org/onlinepubs/9699919799/functions/unsetenv.html>,
> or (if supported) /putenv/() 
> <http://pubs.opengroup.org/onlinepubs/9699919799/functions/putenv.html>
> ...
> 
> IBM is the only vendor I've known to exploit this behavior by copying the
> string value into a volatile buffer.  (Their intent was to spare the
> programmer
> the onus of keeping the argument to putenv() in static storage.)
> This caused
> me problems in Lynx; Tom Dickey generously incorporated my patch to copy the
> returned value.
> 
> But I'm now informed that IBM supplies an alternative, __getenv(), which
> simply returns a pointer to the string the programmer supplied to putenv()
> after the "=".  Might the easy circumvention be:
> 
>     #if defined(__MVS__)
>     #include <stdlib.h>
>     #define getenv __getenv
>     #endif
> 
> ... so getenv is redefined somewhere after any header file mentioning
> getenv but before any call to getenv() in ncurses?

hmm: is there some place where I'm not strdup'ing a long-lived value from
getenv?

-- 
Thomas E. Dickey <address@hidden>
http://invisible-island.net
ftp://invisible-island.net

signature.asc
Description: Digital signature