aboutsummaryrefslogtreecommitdiff
path: root/src/backend/utils/adt/windowfuncs.c
diff options
context:
space:
mode:
authorMichael Paquier <michael@paquier.xyz>2023-01-27 10:19:50 +0900
committerMichael Paquier <michael@paquier.xyz>2023-01-27 10:19:50 +0900
commit783d8abc3b63267194ca21b679caf8d152b93358 (patch)
tree04259c54bea50c7d2e8a6f0ca2c44bd859ea9d90 /src/backend/utils/adt/windowfuncs.c
parent3a28d78089289794fda86cdbd275fc4756c6c6aa (diff)
downloadpostgresql-783d8abc3b63267194ca21b679caf8d152b93358.tar.gz
postgresql-783d8abc3b63267194ca21b679caf8d152b93358.zip
Fix behavior with pg_restore -l and compressed dumps
pg_restore -l has always been able to read the TOC data of a dump even if its binary has no support for compression, for both compressed and uncompressed dumps. 5e73a60 has introduced a backward-incompatible behavior by switching a warning to a hard error in the code path reading the header data of a dump, preventing the TOC items to be listed even if pg_restore -l, with no support for compression, is used on a compressed dump. Most modern systems should have support for zlib, but it can be also possible that somebody relies on the past behavior when copying over a dump where binaries are not built with zlib support (most likely some WIN32 flavors these days, though most environments should provide that). There is no easy way to have a regression test for this pattern, as it requires a mix of dump/restore commands with different compilation options, with and without compression. One possibility I see here would be to have a command-line option that enforces a non-compression check for a build that supports compression, but that does not seem worth the cost, either. Reported-by: Justin Pryzby Author: Georgios Kokolatos Discussion: https://postgr.es/m/20230125180020.GF22427@telsasoft.com
Diffstat (limited to 'src/backend/utils/adt/windowfuncs.c')
0 files changed, 0 insertions, 0 deletions