Re: Getting server crash on Windows when using ICU collation

On Tue, Jun 20, 2017 at 9:06 AM, Ashutosh Sharma <ashu(dot)coek88(at)gmail(dot)com> wrote:
> Hi,
>
> On Tue, Jun 20, 2017 at 2:36 AM, Peter Eisentraut
> <peter(dot)eisentraut(at)2ndquadrant(dot)com> wrote:
>> On 6/19/17 00:42, Ashutosh Sharma wrote:
>>>> If we don't find unconv, isn't it better to fall back to non-UTF8
>>>> version rather than saying command not found?
>>> Well, if any of the ICU package is installed on our system then we
>>> will certainly find uconv command. The only case where we can see such
>>> error is when user doesn't have any of the ICU packages installed on
>>> their system and are somehow trying to perform icu enabled build and
>>> in such case the build configuration has to fail which i think is the
>>> expected behaviour. Anyways, it is not uconv that decides whether we
>>> should fallback to non-UTF8 or not. It's the ICU version that decides
>>> whether to stick to UTF8 or fallback to nonUTF8 version. Thanks.
>>
>> How do we know we're running the uconv command from the installation
>> that we will compile against?
>
> Okay, I think, you are talking about the case where we may have
> multiple ICU versions installed on our system and there might be a
> possibility that the uconv command may get executed from the ICU
> version that we are not trying to link with postgres.
>

To avoid that why can't we use the same icu path for executing uconv
as we are using for linking?

--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com