Make AROS CDVDFS work on AmigaOS (again) #637
Commits on Feb 9, 2024
-
GFLG_GADGHIGHBITS: Fix incorrect usages.
GFLG_GADGHIGHBITS was being used incorrectly in a few locations in Inutition and Icon libraries. Specifically, GFLG_GADGHIGHBITS is a mask against the set of one of the following: GFLG_GADGHCOMP GFLG_GADGHBOX GFLG_GADGHIMAGE GFLG_GADGHNONE
-
Fix for adding windows entry to grub
Execute() returns information whether command was executed, not the return code of command.
-
Adjust to change in keymap limiting Hi to 0x38
Without this change random, unallocated memory was read when using default (coming from keymap.library) keymap and this could cause a crash.
-
Based on tests of several pen-drives formatted under Linux Mint to FAT32 LBA (0x0c)
-
-
-
Allow renaming into exclusively locked directory
Rationale: 1) none of the other file systems does this check 2) it is currently possible to create a file in exclusively locked directory and there is indication that in past it was blocked the same way as renaming. Changing rename synchronize logic in both paths.
-
Pass the tags containing project name and lock to CLI launch
This allows executing tool from "Shell" with an argument via clicking. The code was like this until rev. 31196. It looks like it's a side-effect of that commit, not intention. Note: OS3.9 as well as Scalos both allow this mode This fixes deadwood2/AROS#64
-
Set the correct mask via ohciEnableInt when enabling interrupt
This fix extends rev 39050. When interrupt is disabled via ohicDisableInt is it removed from hc_PCIIntEnMask. Enabling HUBCHANGE interrupt without setting this mask was causing the interrupt to be raised but without code to handle, causing USB to stop issuing next interrupts. Problem became visible when starting ASUS M2N68-AM SE2 with usb keyboard and pendrive connected. USB interrupt was not handled correctly and USB stack was freezing, blocking IRQ5 which also happened to be IRQ of SATA. This fixes deadwood2/AROS#111
-
-
Increase priority of close function to guarantee free() working
At priority 0, C memory management is shutdown.
-
Don't translate error when it is already DOS error
This was causing the previous errno (reported during fixcase) to be reported as effect of Rename rather then error reported in nocase_rename()
-
-
-
-
Field background is used as storage in GetAttr call and thus must be IPTR
-
mgi_Prefs have frame fields of count MUIV_Frame_Count. Having MUIV_Frame_Register as higher value than count was causing reading of random memory in zframe->draw. Note: loading and saving of settings for MUIV_Frame_Register is not implemented. This change is backward compatible. ZunePrefsNew is internal and mgi_Prefs pointer from MUI_GlobalInfo is not even exposed in public headers.
-
Literals are treated as int (32-bit) by default. This causes wrong 64-bit values for negative numbers which are transported via APTR-size fields. Extend the literals to size of long depending on platform.
-
-
Listtree.mcc: Rework CreateImage workaround for 64-bits
The assumption of struct ListImage and NList image structure having obj at the same offset is not valid under 64-bits
-
-
Add workaround for problem visible in C++ code
In C code, p SysBase gives $1 = (struct ExecBase *)addr, however when first C++ debugging information is loaded, this is change into $1 = (ExecBase *)addr. This results in $SysBase = (ExecBase *)addr and gdb complaining about ExecBase type not being known. The hack makes sure that output is always $SysBase = (struct ExecBase*)addr
-
It now support case where address differ in number of digits and falls back to not displaying any images instead of displaying broken text when the hack fails
-
Ignore ('*') comes before width and length, not after width. Previous code was failing where there was * and width. -
Port newer version of random.c from FreeBSD
This version correctly supports 64-bit systems with datasizes fixed at 32 bit integers rather than "longs" as in old version.
-
-
Replace enableusb option with noioapic
'enableusb' is no longer present in source code and 'noioapic' comes handy on some configurations
-
-
Default state of ata_32bit is disabled
If there is no ATA= line in grub command line, it means 32bit switch is disabled.
-
-
-
-
-
-
-
Paths to the font directory should not be translated
-
-
-
CDVDFS: Use CopyMem where possible
Consistent use of CopyMem across the code base
-
CDVDFS: Fix dbprintf / dbinit / dbuninit
dbprintf was changed to pass struct CDVDBase * as the first parameter. This was not consistently fixed in the code base. Function prototypes need to be terminated with a semicolon.
-
CDVDFS: Allow use of OpenLibrary for opening libraries
When compiling CDVDFS for AmigaOS TaggedOpenLibrary is not available
-
CDVDFS: Ensure filesystem IDs are available
On AmigaOS these are not available. Define them conditionally.
-
CDVDFS: Use %lx instead of %p for pointers
RawDoFmt() does not support %p on AmigaOS
-
-
-
-
CDVDFS: Properly cast parameters for FindToolType()
This fixes the following warning: intui.c: In function 'Init_Intui': intui.c:143:24: warning: initialization from incompatible pointer type [-Wincompatible-pointer-types] name = FindToolType (global->g_user_disk_object->do_ToolTypes, "ICONNAME"); ^~~~~~~~~~~~ -
-
-
-
CDVDFS: devsupp: Secure string copies in Get_Startup()
g_resource_fork_extension and g_data_fork_extension are both limited to 17 bytes. Make sure we don't copy more than 17 bytes.
-
CDVDFS: hfs: file names can be 30 characters long
There's no need to make the filename 1 character long and then pad it out with 32 characters.
-
CDVDFS: Use D(bug()); consistently where applicable
CDVDFS has a variety of debug implementations that are inconsistently used across the files. One variant is enough.
-
