On 2026-02-06 Fr 10:11 AM, Tom Lane wrote:
Thomas Munro <[email protected]> writes:
BTW the rest of the patches will reemerge for master, but for the
minimal one back-patched: crake complains about an ABI break due to
GUC table changes. Of course adding a GUC to the stable branches is
unusual and we discussed the need for it in this case. Is that
expected? In what way is it part of the ABI? How would one determine
in advance that the ABI checker will complain?
We have very little experience so far with libabigail, so there's
not any store of knowledge hereabouts on what it'll complain about.
The message is complaining that sizeof(ConfigureNamesEnum) changed,
which it did, but I don't see how that value would be visible to
external modules. So maybe a bug in libabigail?
I guess because of this in guc_tables.h:
extern PGDLLIMPORT struct config_generic ConfigureNames[];
cheers
andrew
--
Andrew Dunstan
EDB: https://www.enterprisedb.com
