[x] Seems to be an accidental double prefixing with session_dir then before loading.
[x] SessionManager::open still has it right: test/user/session\god wars.save
[x] Maybe load_snapshot() does the prefixing on its own, which works with quick-load names, but not with sessions? Or is it that the session loader just shouldn't do the prefixing but leave it to the "raw" loader logic instead (I don't thing so!)?
[x] SessionManager::open has it still right even after the load error (Failed to load session state (from test/user/session\god wars.save)!), so it's definitely in the load logic!
[x] Yes... In SimApp::load_snapshot:
string fname = sz::prefix_if_rel(cfg.session_dir, unsanitized_filename);
[x] But then how come the regression tests can still load sessions?! :-o
[x] Well, the test scripts seem to use abs. paths:
Szim::SessionManager::open C:\sz\prj\Out_of_Nothing\test\regression\_baseline-d5de5369/1000_bodies-START.state
[x] OK, but then again: how come that the same problem doesn't seem to exist with SessionManager::close(), even though filenames are also prefixed by save_snapshot(), symmetrically?
-> Hmm, it's because the session close logic doesn't do prefixing if a name is set with --session=name! :-o
[x] But --session-save-as=name does, so that should fail! (It does...)
[x] And so does the default naming branch (--session) , so that should fail, too! (And it does, too!)
(Possibly a regression.)
or (just to confirm that it has nothing to do with other options etc.):
->
[x]
session_dir
istest/user/session
: OK[x] Seems to be an accidental double prefixing with
session_dir
then before loading.SessionManager::open
still has it right:test/user/session\god wars.save
load_snapshot()
does the prefixing on its own, which works with quick-load names, but not with sessions? Or is it that the session loader just shouldn't do the prefixing but leave it to the "raw" loader logic instead (I don't thing so!)?SessionManager::open
has it still right even after the load error (Failed to load session state (from test/user/session\god wars.save)!
), so it's definitely in the load logic!SimApp::load_snapshot
:string fname = sz::prefix_if_rel(cfg.session_dir, unsanitized_filename);
[x] But then how come the regression tests can still load sessions?! :-o
Szim::SessionManager::open C:\sz\prj\Out_of_Nothing\test\regression\_baseline-d5de5369/1000_bodies-START.state
[x] OK, but then again: how come that the same problem doesn't seem to exist with
SessionManager::close()
, even though filenames are also prefixed bysave_snapshot()
, symmetrically? -> Hmm, it's because the session close logic doesn't do prefixing if a name is set with--session=name
! :-o--session-save-as=name
does, so that should fail! (It does...)--session
) , so that should fail, too! (And it does, too!)