Am 11.07.2016 um 20:14 schrieb Bernhard Nortmann:
This patch introduces a new "export" environment operation (env_op_export) and the corresponding access flag ENV_FLAGS_VARACCESS_PREVENT_EXPORT; so that env_flags_validate() may now check requests to export specific variables.
In turn, hexport_r() makes uses of this ability to suppress the export of variables that are flagged accordingly.
Note that env_flags_validate() and hexport_r() will respect H_FORCE and H_PROGRAMMATIC flags, allowing to bypass the export filtering. H_PROGRAMMATIC gets used within env_print() to make sure all variables are listed. This is necessary because env_print() is essentially an "export to text" operation.
Signed-off-by: Bernhard Nortmann bernhard.nortmann@web.de
[...]
diff --git a/lib/hashtable.c b/lib/hashtable.c index 02b4105..708319d 100644 --- a/lib/hashtable.c +++ b/lib/hashtable.c @@ -621,6 +621,10 @@ ssize_t hexport_r(struct hsearch_data *htab, const char sep, int flag, if ((flag & H_HIDE_DOT) && ep->key[0] == '.') continue;
if (env_flags_validate(ep, NULL, env_op_export,flag & H_FORCE) != 0)continue;list[n++] = ep; totlen += strlen(ep->key) + 2;
This is also a remnant of previous experiments, and a too narrow condition that doesn't match the commit message's description. If "flag" is to be fully respected (H_FORCE | H_PROGRAMMATIC bits), then the H_FORCE mask (bitwise and operation) needs to be removed. Fixed in my local branch.
Regards, B. Nortmann