to make koreader on Android more stable
and with these opt params:
```
require("jit.opt").start("sizemcode=64","maxmcode=64", "hotloop=10000")
```
The strategy here is that we only use precious mcode memory (jitting)
on deep loops like the several blitting methods in blitbuffer.lua and
the pixel-copying methods in mupdf.lua. So that a small amount of mcode
memory (64KB) allocated when koreader is launched in the android.lua
is enough for the program and it won't need to jit other parts of lua
code and thus won't allocate mcode memory any more which by our
observation will be harder and harder as we run koreader.
Because we cannot deal with it properly...
We'd need to be able to stop eating all input, and have a lot of luck
with refresh timings to actually have the popup visible at the right
time.
TL;DR: it's a mess, kill it with fire.
Fix#1811
Which I broke the last time I touched this.
NOTE: I'm not quire sure why the bbsave/bbrestore on stop/resume did a
double check (screensaver & charging), because that effectiely disabled
it when charging, which I don't quite understand...
Also makes sure the needsScreenRefreshAfterResume flag is honored on
Kindles, because we need it on FW >= 5.7.2
Note that this *breaks* the behavior with passcode enabled, which was
unwittingly fixed in said previous changes...
re #1811
This workaround makes a reservation of mmap slots that are hopefully
near the address of `mcode_alloc_at`. Then it unmaps those slots and
lets luajit allocate a large slot for mcode.
When launching KOReader with the framework up from KUAL...
Our gentle hide/unhide method doesn't work anymore, which leads to
various issues, the most obvious being not getting a refresh to the
default UI on exit...
Currently only tested on Ubuntu-touch emulator with framework
ubuntu-sdk-14.10 for armhf.
The ubuntu-touch port is binary compatible with the Kobo port
major changes in this PR are:
1. rename the emulator device to sdl device since both the emulator
and the ubuntu-touch target use libsdl to handle input/output.
2. ubuntu-touch app has no write access to the installation dir so
all write-outs should be in a seperate dir definded in `datastorage`.
Properly refresh the screen when exiting with the framework running.
The lua code to handle that is called while cvm is SIGSTOPP'ed, so it
doesn't help.
To avoid potential issues with USBMS.
Namely, make sure CWD doesn't point to the userstore for persitent
processes.
Make sure wpa_supplicant doesn't try to use our own OpenSSL lib, which,
again, is in the userstore area.
Properly kill our dhcp client invocation on when killing WiFi
Also, some related fixes to the startup script:
Properly kill wpa_supplicant & the dhcp client when we kill
wifi before rebooting to nickel.
On kindle, kobo and pocketbook the data directory is the current
running directory but on Android the app is installed in system
defined location and users may have no access to that location.
The same circumstances should be true for the upcoming Koreader for
Ubuntu touch, so the data directory (in which tessdata, dictionaries,
global settings, persistant defaults and probably history data are
stored) could be stored in another place.
Thanks to @Markismus's questions, I realized that some of what i thought
was true, wasn't.
First, pkill is a terrible idea to check for interpreted scripts.
Second, pidof is also potentially not that great for interpreted
scripts, because it'll only work with a shebang, and one that is
following the Linux syntax.
We don't have the full version with the -x to paper over that,
so use pgrep -f instead.
cf.
a736a571d2 (commitcomment-10948910)
for the gory details.
pkill probably only checks the actual name of the binary.
We were checking a shell script, which I'm guessing is not launched
via a shebang, so that name was actually an arg to sh, or something like
that.
Use pidof instead, which matches the full command line.
For some unknowable reason, restarting nickel on exit deadlocks on my
H2O, so, do a full reboot instead, at least *that* works.
I'm probably one of the only people running KO manually anyway...
Compatibility with KSM is resolved. Koreader will now work from any directory it is installed in.
KSM7 tweaks `koreader.sh` line 5 from a hardcoded path to the line used above. I've added the change in line 12 to allow OTA updating.
Anyone who encounters random freeze of koreader for Android
is encouraged to enable this by adding a file at
"/sdcard/koreader/patch.lua" with the content of:
```
require("jit.opt").start("sizemcode=256","maxmcode=256")
for i=1,100 do end -- Force allocation of one large segment
```
The sizemcode and maxmcode could be adjusted from 196 to 512 to
make koreader best fit on your Android device.
The commands in networkmgr.lua rely on an assigned value of PLATFORM. As discussed in #1423 this is only the case for child processes of rcS. This seems to be a fine place to check the value of PLATFORM.
It turns out that one of our mupdf patch reads the FONTDIR env var
and uses it in a totally different way so we will use another env var
EXT_FONT_DIR to define external font directory for different platforms.
to support custom font directory for EPUB documents
Now Koreader could find fonts in the "fonts" directory in the USB root
directory of kindle, Kobo and PocketBook devices, thus no need to copy
fonts to "koreader/fonts" directory.
