...strictly speaking, what we use isn't correct (eg we should use es_MX
instead of es-mx) but the latter is clearer on the download page.
Change-Id: Ia1a21c4259826ca2a0f6755df1cb3db94e4a50b3
We expect a fixed FRAMEBUFFER_SIZE that's width*height*bitdepth, and we
mmap()ed that in. However, when doing the initial fb clear, we
were using the hardware-provided 'finfo.smem_len' which could be
larger than FRAMEBUFFER_SIZE. This overran our mmap and triggered
a segfault.
Correct this by mmaping (and clearing) the entire smem_len. As a safety
measure, panic if smem_len is smaller than our expected FRAMEBUFFER_SIZE
Change-Id: I3222139c7aed6e8e8ee232b1730edd5cd70065ff
GPT superblocks are located at sector 1 and max_sector-1. If the system
uses a "virtual" sector that's larger than the drive's logical sector,
we need to map those virtual sector numbers to the drive's logical
sector.
If DEFAULT_VIRT_SECTOR_SIZE is defined, try that multiplier as well
as the standard multiplier of 1.
It's not practical to try every intermediate value so instead, if
DEFAULT_VIRT_SECTOR_SIZE is defined, try that as well as the standard
multiplier of 1.
This still leaves a handful of targets that don't set DEFAULT_VIRT
but do set MAX_VIRT.
Change-Id: I3accffcb97436b043836e072bfc620318a9b1230
Basically the GPT is supposed to live at sector 1, but a backup copy is
stored on the final sector.
This gives us a little bit of extra flexibility on systems that might
require sector 1 for other things, but in any case it's a more robust
arrangement.
Change-Id: I8925ffc743629cf2eba51861042492e35b41664b
A function that returned bool was an alias for a function that returned
an int. While that original function was limited to returning either 0
or 1, this aliasing is technially a no-no. So create a small shim to
make the warning go away.
Change-Id: I4d7807730234928bd59d75f13a4e4adeabbf655e
12b9419006 stripped out leading/trailing whitespace, but the Latvian
translation used a blank string for VOICE_OH which got wiped out.
This string is only used when voicing the time, eg: 12:05 is spoken in
English as "twelve oh five". Doing a little digging into Latvian, it
seems that for 24 hour clocks it would be voiced as "<twelve> nulle <five>"
instead, so I changed VOICE_OH to that.
Now for 12-hour clocks, it seems Latvian would voice "12:05" as "<five
past twelve>" but we don't currently have the ability to voice things
in this manner. Will be addressed in a future patch.
Change-Id: I2c1bcd2e9caff508de3199da29c2e994e9d00c3c
We normally create a table of the partition sizes/types present
on a drive. Howeever, if the drive is set up as a "superfloppy",
where there is no partition table and a single filesystem starting
at sector 0, this "pinfo" table is not populated.
So now, populate the pinfo table with a single entry that matches
the filesystem type, start, and size.
Change-Id: Ifa8760909109d67ff96481b1fc7f26c64280a00a
Languages:
* Get rid of leading space on LANG_ID3_VBR [ " (VBR)" ]
* Fix up sole user to insert the space programmically
updatelang:
* strip leading and trailing spaces on all phrases except VOICE_PAUSE
voice.pl
* Debug logging with UTF-8 output
* Explicitly delete tab character from voiced strings
Change-Id: Ie466793479ce15ce7a9553770583a070530e7afd
All toolchain dependencies are circa the GCC 9.5.0 release:
GCC 9.5.0, binutils 2.38, gmp 6.2.1, mpfr 4.1.1, mpc 1.2.1, isl 0.18
Native:
* arm - mini2g, nano2g seems to work
- ipod6g boot hang on 9.5.0 (possibly leading to disk corruption)
* mips - xduoox3 seems stable so far
* m68k - binaries untested
Hosted:
* arm - binaries untested
* mips - erosq seems stable so far
* Android NDK - unchanged at GCC 4.9.4
Change-Id: I4ca00edd49e1936b32f68f1e7b34df38e1d1d76e
* binutils (mips) uses its own 'static_assert'
* linux (mips) uses constexpr when generating headers
With these patches, all toolchains build successfully under GCC15
(These patches are still needed with the pending GCC9.5 bump)
Change-Id: Id2f8cca414d2907b6aa29612ae4fb0e8b5f79f79
All translations with at least 80% coverage are upated.
Most of these also got a few mechanical corrections as well.
Change-Id: Id4cf5e6c066cb7cd0a73bcd5ad68135c9aeb3294
The Echo R1 is a new open-hardware music player design, based
on the STM32H743 microcontroller. Schematics and hardware
documentation for it can be found here:
- https://github.com/amachronic/echoplayer
This is an incomplete port. The bootloader can be loaded using
OpenOCD and it can draw to the LCD using SPI. SDRAM is working
but hasn't been extensively tested.
Change-Id: Ifd2bee15c49868fbc989683d3ca14dce48bf3e18
The bootloaders for version 3 and 4 hardware used separate target
IDs which now overlap with the Nano 3G / Nano 4G ports. It likely
makes sense for all the ErosQ native builds to share the same ID.
Change-Id: I40d166758744e8ba25931e568da11f374b2d7b90
The "fix" produces different asm and I don't have the bandwidth
to determine if it's functionally equivalent.
Change-Id: I822557d1244f8e8a05e7ebfc45da2643152345a4
/* offset = -3db * (ratio - 1) / ratio */
db_curve[2].offset = (int32_t)((long long)(-3 << 16)
* (ratio - 1) / ratio);
Results in a warning: left shift of negative value [-Wshift-negative-value]
To fix this, replace the '-3' with -3UL.
(As that is effectively what is already happening)
Change-Id: I5aa269ff332703d3c2d889fa032d2fdc403ff14f
