masaofujii / pg_cheat_funcs Goto Github PK

Hi Fujii-san,

I am thinking about writing a pg_cheat_funcs() patch to call the function DropDatabaseBuffers(). I would use something like this for testing my corruption detection tools. I can modify the on-disk representation myself, possibly by using my pg_hexedit tool, and then call a Postgres function to invalidate anything that happens to be in shared_buffers immediately. I can then test some amcheck function, or whatever it happens to be, to make sure it detects the issue. This new function would be a convenience thing for this very specific kind of work.

I was going to just write a PR, but thought I should ask first: If I wrote a function like this, would you accept it? It's not obvious that you would, because calling DropDatabaseBuffers() outside of something like a DROP DATABASE is generally very dangerous. The function drops dirty buffers without cleaning them/writing them out first. I thought about just making the function do a checkpoint first, which would be easy, but quickly realized that that would be even more dangerous: it would lead to a false sense of security, because it would only work 99% of the time. (There'd still be race conditions)

What do you think? If it's not suitable, I can just do something on my own.

Thanks!

Hi Fujii,

Thanks for making pg_cheat_funcs!

make installcheck fails on Ubuntu 20.04.2 LTS running Ubuntu 13.2-1.pgdg20.04+1.

Maybe I need to run some additional apt-get command to install the EUC_JP encoding before running the tests?

$ cat regression.diffs
diff -U3 /home/joel/postgresql-extensions/pg_cheat_funcs/expected/pg_eucjp.out /home/joel/postgresql-extensions/pg_cheat_funcs/results/pg_eucjp.out
--- /home/joel/postgresql-extensions/pg_cheat_funcs/expected/pg_eucjp.out	2021-04-26 06:55:53.064946839 +0000
+++ /home/joel/postgresql-extensions/pg_cheat_funcs/results/pg_eucjp.out	2021-04-26 07:00:01.184044549 +0000
@@ -1,17974 +1,5 @@
CREATE DATABASE regtest_cheat_funcs_eucjp ENCODING 'EUC_JP' TEMPLATE template0;
+ERROR:  encoding "EUC_JP" does not match locale "en_US.UTF-8"
+DETAIL:  The chosen LC_CTYPE setting requires encoding "UTF8".