diff options
author | Michael Paquier <michael@paquier.xyz> | 2023-03-15 12:55:51 +0900 |
---|---|---|
committer | Michael Paquier <michael@paquier.xyz> | 2023-03-15 12:55:51 +0900 |
commit | 765f5df726918bcdcfd16bcc5418e48663d1dd59 (patch) | |
tree | 441d40c3a43cd6cf1ddcf69cbf235d45add8f620 /src/backend/parser/parse_expr.c | |
parent | 89e46da5e511a6970e26a020f265c9fb4b72b1d2 (diff) | |
download | postgresql-765f5df726918bcdcfd16bcc5418e48663d1dd59.tar.gz postgresql-765f5df726918bcdcfd16bcc5418e48663d1dd59.zip |
Improve WIN32 port of fstat() to detect more file types
The current implementation of _pgfstat64() is ineffective in detecting a
terminal handle or an anonymous named pipe. This commit improves our
port of fstat() to detect more efficiently such cases by relying on
GetFileType(), and returning more correct data when the type found is
either a FILE_TYPE_PIPE (_S_IFIFO) or a FILE_TYPE_CHAR (_S_IFCHR).
This is part of a more global fix to address failures when feeding the
output generated by pg_dump to pg_restore through a pipe, for example,
but not all of it. We are also going to need to do something about
fseek() and ftello() which are not reliable on WIN32 for the same cases
where fstat() was incorrect. Fixing fstat() is independent of the rest,
though, which is why both fixes are handled separately, and this is the
first part of it.
Reported-by: Daniel Watzinger
Author: Daniel Watzinger, Juan José SantamarÃa Flecha
Discussion: https://postgr.es/m/b1448cd7-871e-20e3-8398-895e2d1d3bf9@gmail.com
Backpatch-through: 14
Diffstat (limited to 'src/backend/parser/parse_expr.c')
0 files changed, 0 insertions, 0 deletions