I wish I had realized this the last time it came up - it would have saved me a lot of trouble.
But as it turns out, the more recent travelling code makes all of this completely unnecessary, working perfectly fine with deleting the player pawns along with the rest of the thinkers before loading the stored ones from the savegame (and getting rid of those in G_FinishTravel.)
And with a sane savegame format that does not depend on side effects from how the thinker serializing handled linking into the lists the old code was even harmful, leaving voodoo dolls behind.
I had the exact same effect when I tried to reshuffle some things for reliably restoring portals, but did not make the connection to interference between two mutually incompatible player travelling mechanisms that just worked by sheer happenstance with the original order of things.
- changed S_GetMusic to return a const pointer to the actual music name instead of a copy. The only thing this is used for is the savegame code and it has no use for a copy, it can work far more efficiently with a const pointer.
After testing with a savegame on ZDCMP2 which is probably the largest map in existence, timing both methods resulted in a speed difference of less than 40 ms (70 vs 110 ms for reading all sectory, linedefs, sidedefs and objects).
This compares to an overall restoration time, including reloading the level, precaching all textures and setting everything up, of approx. 1.2 s, meaning an increase of 3% of the entire reloading time.
That's simply not worth all the negative side effects that may happen with a method that highly depends on proper code construction.
On the other hand, using random access means that a savegame version change is only needed now when the semantics of a field change, but not if some get added or deleted.
- do not I_Error out in the serializer unless caused by a programming error.
It is better to let the serializer finish, collect all the errors and I_Error out when the game is known to be in a stable enough state to allow unwinding.
It turned out this may not be done automatically when opening the savegame - it has to be done later, after the pre-spawned map thinkers and all connected objects have been destroyed.
The object deserializer also has to be rather careful about dealing with parse errors, because if something goes wrong a whole batch of uninitialized or partially initialized objects will be left behind to destroy.
This means that no object class may assume that anything but the default constructor has been run on it and needs to check any variable it may reference.
- always make the top level object randomaccess when opening a JSON file for reading. Some things won't work right if this is opened for sequential access.
- Several mods were able to just take advantage of A_SetRipperLevel and the likes, essentially bypassing this gate so there really is no point in doing this anymore.
This is so that PNGs can be written to memory, not just to an external file. stdio's FILE cannot be easily redirected but a C++ class can.
The writer is very simple and primitive right now, allowing no seeking, but for the job at hand it is sufficient.
Note that large parts of savegame creation have been disabled, because they are about to be rewritten and it makes no sense to adjust them all before.
- converted sound and canvas texture serialization.
- refactored file_zip, so that it can be used to load loose zip files and extract their compressed data directly.
- added handling to FSerializer to generate and consume compressed Zip file entries.
If all goes well this will allow saving savegames as Zips when the rework is done, which will make analyzing them a lot easier.
