While trying to debug a recent bug report on hash indexes [1], I noticed that pg_verify_checksums don't work on Windows (or at least in my environment).
initdb -k ..\..\data pg_verify_checksums.exe ..\..\Data pg_verify_checksums: short read of block 0 in file "..\..\Data/global/1136", got only 15 bytes
I have debugged and found that below code is the culprit.
scan_file(char *fn, int segmentno) { .. f = open(fn, 0); .. int r = read(f, buf, BLCKSZ);
if (r == 0) break;
if (r != BLCKSZ) { fprintf(stderr, _("%s: short read of block %d in file \"%s\", got only %d bytes\n"), progname, blockno, fn, r); exit(1); } .. }
We are opening the file in text mode and trying to read the BLCKSZ bytes, however, if there is any Control-Z char, it is treated as EOF. This problem has been mentioned in the comments in c.h as follows: /* * NOTE: this is also used for opening text files. * WIN32 treats Control-Z as EOF in files opened in text mode. * Therefore, we open files in binary mode on Win32 so we can read * literal control-Z. The other affect is that we see CRLF, but * that is OK because we can already handle those cleanly. */
So, I think we need to open the file in binary mode as in other parts of the code. The attached patch fixes the problem for me.
Thoughts?
Yikes. Yes, I believe you are correct, and that looks like the correct fix.
I wonder why this was not caught on the buildfarm. We do have regression tests for it, AFAIK? Or maybe we just lucked out there because there was no ^Z char in the files there?
