Re: Patch: Implement failover on libpq connect level.

On Thu, Oct 27, 2016 at 9:46 AM, Mithun Cy <mithun.cy@enterprisedb.com> wrote:
> On Wed, Oct 26, 2016 at 8:49 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>> Let me know your thoughts.
>
> One small issue. I tried to run make installcheck after applying patch there
> seems to be a invalid write access in code (resulting in crash).
>
> ==118997== Invalid write of size 1
> ==118997==    at 0x4E3DDF1: connectOptions2 (fe-connect.c:884)
> ==118997==    by 0x4E3D6FF: PQconnectStartParams (fe-connect.c:608)
> ==118997==    by 0x4E3D553: PQconnectdbParams (fe-connect.c:465)
> ==118997==    by 0x413067: main (startup.c:245)
> ==118997==  Address 0x5dc4014 is 0 bytes after a block of size 4 alloc'd
> ==118997==    at 0x4C29BFD: malloc (in
> /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
> ==118997==    by 0x4E3DD3C: connectOptions2 (fe-connect.c:880)
> ==118997==    by 0x4E3D6FF: PQconnectStartParams (fe-connect.c:608)
> ==118997==    by 0x4E3D553: PQconnectdbParams (fe-connect.c:465)
> ==118997==    by 0x413067: main (startup.c:245)
>
> After locally fixing this by allocating an extra space for string terminal
> character. make installcheck  tests pass.
> -(char *) malloc(sizeof(char) * (e - s));
> + (char *) malloc(sizeof(char) * (e - s + 1));

Thanks.  Here's a new version with a fix for that issue and also a fix
for PQconnectionNeedsPassword(), which was busted in v1.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company