Fix memleaks in http.c and encode.c

[rikylescak]

Hi,
there are two memory leaks found by valgrind in cups:

=25357== 41 bytes in 2 blocks are definitely lost in loss record 678 of 1,250
==25357== at 0x484186F: malloc (vg_replace_malloc.c:381)
==25357== by 0x4DB7FBE: strdup (strdup.c:42)
==25357== by 0x4CB01EF: http_add_field (http.c:3676)
==25357== by 0x4CB0E11: _httpUpdate (http.c:2891)
==25357== by 0x4CB0EBA: UnknownInlinedFun (http.c:2948)
==25357== by 0x4CB0EBA: httpUpdate (http.c:2918)
==25357== by 0x4CB25A8: http_tls_upgrade (http.c:4620)
==25357== by 0x4CC7A07: cupsGetResponse (request.c:438)
==25357== by 0x4CCBD4A: cupsDoIORequest (request.c:232)
==25357== by 0x125875: browse_poll_cancel_subscription (cups-browsed.c:11173)
==25357== by 0x11339F: main (cups-browsed.c:12784)

==25357== 480 (72 direct, 408 indirect) bytes in 1 blocks are definitely lost in loss record 1,204 of 1,250
==25357== at 0x4846464: calloc (vg_replace_malloc.c:1328)
==25357== by 0x4CB7581: ippNew (ipp.c:2622)
==25357== by 0x4CA5533: _cupsEncodeOption (encode.c:644)
==25357== by 0x4CA67D7: UnknownInlinedFun (encode.c:851)
==25357== by 0x4CA67D7: cupsEncodeOptions2 (encode.c:725)
==25357== by 0x122C0E: update_cups_queues (cups-browsed.c:8813)
==25357== by 0x4B0E980: g_timeout_dispatch (gmain.c:4933)
==25357== by 0x4B0E12F: UnknownInlinedFun (gmain.c:3381)
==25357== by 0x4B0E12F: g_main_context_dispatch (gmain.c:4099)
==25357== by 0x4B63207: g_main_context_iterate.constprop.0 (gmain.c:4175)
==25357== by 0x4B0D852: g_main_loop_run (gmain.c:4373)
==25357== by 0x11324C: main (cups-browsed.c:12747)

[zdohnal]

Hi @rikylescak ,

thank you for the PR!

Freeing the IPP structure collection looks fine here, use counter will prevent the complete free of the pointer and ippDelete() calls happen after ippSetCollection() calls on other places as well.
Memory allocation in _httpUpdate() happens when there is a known field in HTTP header during upgrading the connection - if httpClearFields() doesn't happen, memcpy() will rewrite the fields during restoring original values, so we lose the pointer.

LGTM, I'll start the test suite.

[michaelrsweet]

LGTM

[zdohnal]

Merged as commit d992418 , thank you for the contribution!