On 06.10.23 02:24, Krishnakumar R wrote:
>> elog(INFO, "Open bki file %s\n", bki_file);
>> + boot_yyin = fopen(bki_file, "r");
>>
>> Why is this needed? It already reads the bki file from stdin?
> We no longer open the bki file in initdb and pass to postgres to parse
> from stdin, instead we open the bki file directly in bootstrap and
> pass the file stream to the parser. Hence the need to switch the yyin.
> Have added a comment in the commit logs to capture this.
Why this change? I mean, there is nothing wrong with it, but I don't
follow how changing from reading from stdin to reading from a named file
is related to moving the parameter substitution from initdb to the backend.
One effect of this is that we would now have two different ways initdb
interacts with the backend. In bootstrap mode, it reads from a named
file, and the second run (the one that loads the system views etc.)
reads from stdin. It's already confusing enough, so any further
divergence should be adequately explained.