On 02/16/2018 04:04 PM, Mats Wichmann wrote:
Just to get some clarity on this ...
Yesterday I proposed fixes for a few buffer size problems reported by
static analysis (valgrind).
In commenting, George Nash noted that in some cases I had added
oic_string.h in order to use OICStrcpy, but that oic_string.h is not
part of the public API.
good valid point
So just the three questions this time:
1. is that intentional? oic_string.h definitely not part of the API?
well It's not my part, but I suppose it was created to track internal allocs
does it make sense to do mix applications allocations and library ones ?
if so then the header should be exposed as public or experimental ?
To be established at:
https://wiki.iotivity.org/api
2. it is true that samples should only use public APIs? (I found
multiple instances that do use oic_string.h, did not check for anything
else)
AFAIK, some are using experimental things, planned to become public
3. if (2) is true, do we need an "api checker" to run on the samples to
make sure they're clean?
This would help for sure,
an other way would be to build app using installed headers (scons
install) only
and review each of them to make only public APIs are exposed.
Note, I have pending changes on this (loggers etc)
I promised that I'll help on it,
once all 1.3-rel pending changes are merged to master
and infra is back on track
( https://jira.iotivity.org/browse/IOT-2279 )
My 2c
--
mailto:philippe.co...@osg.samsung.com gpg:0x467094BC
https://blogs.s-osg.org/author/pcoval/
_______________________________________________
iotivity-dev mailing list
iotivity-dev@lists.iotivity.org
https://lists.iotivity.org/mailman/listinfo/iotivity-dev
