Daniel Shahaf wrote on Sat, Feb 19, 2022 at 13:35:22 +0000: > Julian Foad wrote on Fri, Feb 18, 2022 at 17:00:02 +0000: > > I committed the proposed fix in r1898184: > > [[[ > > On the 'pristines-on-demand-on-mwf' branch: fix authz_tests.py 31. > > > > This test runs an 'update' after read access permission has been removed > > from a repository path. The text base sync code errored out when it > > tried to download this text base. > > > > We fix it by making the text base sync code ignore authorization errors, > > leaving any such base unfetched and continuing to try the others. > > ... > > * subversion/libsvn_wc/wc_db_textbase.c > > (svn_wc__db_textbase_sync): If read access is unauthorized, ignore the > > error and continue even though we failed to fetch the textbase. > > ]]] > > This seems to have regressed the failure mode of «svn cat iota@BASE» > when iota is locally-modified and has no read access.
Should iota be set to server-excluded? Presumably, if we do that, later code won't try to access iota's pristine? Also, the test failed because of an SVN_ERR_RA_NOT_AUTHORIZED error from the update at the end. Is that actually a bug? Or perhaps it's actually the/a right thing to do? What are the expected semantics of working copies that _already_ contain versioned dirents (file contents and properties) that are no longer readable to them? Cheers, Daniel
