@Weedy_Weed_Smoker Using the core with dynarec i got in Hercules to the transition between levels 2 and 3(the game always freezes there) and took a snapshot. Then I booted from a snapshot on the core without dynarec r19 and managed to load level 3! 
Since this was the most difficult place(there was a freeze in it with any BIOS), then all other freezes most likely will not be there anymore, but I can reliably check this only with graphics acceleration.
@Weedy_Weed_Smoker I do not fully understand what the numbers after ReARMed the version number mean (example: r22 000000) - perhaps this is a unique build number, or maybe part of a minor version(then it might be useful).
Anyway. I tested several core versions. Here are the results:
- r19 no series (interpreter) / works perfect, but slow.
- r22 no series (interpreter) / works perfect, but slow.
- r22 no series (dynarec) / application crashes.
- r22 15cfb55 (dynarec) / freezes and not save, but fast.
- r22 d56340b (dynarec) / freezes and not save, but fast.
- r22 b715d67 (dynarec) / freezes and not save, but fast.
Sources:
I also tried to disable the dynarec in your version(r22 d56340b, r22 b715d67 with Neon and PEOpS) in the core settings, as expected, it slowed down the game, but the error did not fix it.
@dagazik Ok, I compiled a version dating back from february 2020 to test, can you update the core?
@Weedy_Weed_Smoker Thanks for the compilation, but unfortunately the same problem: your build of ReARMed r22 no series (dynarec) from February 2020 - freezes and not save, but fast. 
I think that in order to get closer to the solution, you need to make 2 builds r19 from here(dynarec and interpreter): https://notaz.gp2x.de/cgi-bin/gitweb.cgi?p=pcsx_rearmed.git
Here is a direct link to the source code r19: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?p=pcsx_rearmed.git;a=snapshot;h=101e053c72e9299bfbd1cbf6a8de9ba6cff11e4c;sf=tgz
This action will give a complete understanding of what the problem is.
But it is already clear that this seems to be happening only for old IOS. Better in testing, I will download from the http-link than from the repository(so as not to create inconvenience to other users).
@dagazik I uploaded a new test version from jul 2018 on the buildbot, please try to update and report if that works!
Don’t worry, this is the old iOS versions’ buildbot, people using it will have the same problem if we don’t fix it…
@Weedy_Weed_Smoker As I answered in a nearby thread: RA: PCSX ReARMed crashing on Final Fantasy 7 - ios 9.35
Your build of ReARMed r22 4adc2c29 from jul 2018 - freezes not saved and also slows… In my case, this version has collected all the worst 
@Weedy_Weed_Smoker Based on the data from the FF7 + ios9 branch: can conclude that me need a core no older than May 2015(2015 - 2013), but built with iOS5 support.
@dagazik I compiled a version from March 2016 (which is only 4 commits after the one you linked to…)
Commits on Mar 5, 2016
large file support
Commits on Feb 29, 2016
psxmem: use rounding that's more likely to work
drc: handle any immediate
Commits on Jan 8, 2016
menu: fix input config for device names ending with space
You can update the core!
I would have loved to have an usable iOS5/9 device to track the exact commit that broke older iOS to fix it on the latest version, but at least it will (hopefully) work, now!
I can confirm that its working on ios 9.3.5. If you want I can try all the commits and report back on the exact one that breaks!
@AfonsoH Don’t worry, I had to compile on an old macOS version (OS X 10.9) and it’s kinda tedious…
I will do that when I repair my iPhone 4s’ screen, it’s jailbroken on iOS 6 so I could test that, and I can eventually update it to iOS 9!
Thanks for all the efforts! If there is any way I can help with any testing on ARM-32bit-iOS9 related just tell me
@Weedy_Weed_Smoker Bad work again in my case with iOS5(version from March 2016). 
But this time the behavior has changed: the application crashes.
@Weedy_Weed_Smoker Thanks a lot for all attempts. Unfortunately, RetroArch + ReARMed from December 2014(r20) app crashes. (it seems that it will not work on iOS5 due to some aspect of compatibility). 
We have the following results ReARMed:
- r22 4adc2c29 from Jul 2018 and newer(dynarec) / freezes and not save, but fast.
- r22 no series from March 2016 and older(dynarec) / application crashes.
- r19 no series date unknown (interpreter) / works perfect, but slow.
- r22 no series from May 2015 (interpreter) / works perfect, but slow.
The first two edition in list - are your builds.
Perhaps this is somehow related to this topic: https://github.com/notaz/pcsx_rearmed/issues/23
@Weedy_Weed_Smoker Сan try to compile the version “r20 from December 2014” without any acceleration at all(as described on the forum at the link above).
If this time the problem is related to the dynarec code - it error will be identified.
Then just add a couple of fixes to the code as shown here:
@dagazik I compiled 2 variants of pcsx_rearmed r20, one that disables dynamic recompiler (dynarec), the “_interpreter” version, and another with GPU set as peops instead of neon…
I tried compiling without neon optimizations but it didn’t even compile…
(Download from MediaFire, I didn’t bother to update the buildbot)
@Weedy_Weed_Smoker I tested it, here is the result:
- ReARMed interpreter version “r20 from December 2014” / works perfect!!! Unloaded from RAM, loads save from snapshots, and loads those games that previously freezes!
- ReARMed GPU set as peops / application crashes.
We must try to fix the code and compile with dynarec/neon (it seems we are one step away from success).
The fix code is here!
@dagazik I compiled another version from December 2013, the fix you linked to is not affecting iOS but generic arm platforms…
Similar behavior(r19 version from December 2013) - application crashes.
Application to crash from version ReARMed March 2016 and older(we tested 2016 - 2013).
