Allow toggling CRe's new dithering & scaling #888

NiLuJe · 2019-04-14T18:42:35Z

Plus, implement said dithering in our own BB code, for ImageWidget

NiLuJe · 2019-04-15T18:15:44Z

Oh, joy, even GCC 5.4 defaults to gnu++98 -_-"

Frenzie · 2019-04-15T20:16:12Z

Makefile.defs

-# Anything below GCC 6 gets to be forcefully switched to decent standards...
-# NOTE: We're also assuming Clang >= 3.2, but that should more or less hold true on our end.
+# Anything below GCC 6 gets to be forcefully switched to decent standards, because some of our deps require C11 & C++11 support.
+# NOTE: Technically, this also means we require Clang >= 3.2, but that should more or less hold true on our end, because that's truly ancient.


I don't think older than Clang 5 realistically occurs for us in any scenario.

Yeah, that's what I thought. Even the Travis mention a few lines before mentions 3.4, and that's already an old comment, for an old Travis image ;).

it anyway

Let's see PB's GCC choke on C11/C++11...

Here's hoping that'll do the trick even on PB...

(IIRC, GCC 6.0 never came out, so, eh).

It's calling CXX w/ CFLAGS, which is braindead on all kinds of levels.

Anbdroid TCs are stuck on GCC 4.9 (Pre-GPL3).

test/rsa_complex.c:17:23: fatal error: complex.h: No such file or directory

Frenzie · 2019-04-18T17:48:12Z

thirdparty/openssl/CMakeLists.txt

@@ -62,7 +62,7 @@ set(PATCH_CMD1 sh -c "patch -N -p1 < ${CMAKE_CURRENT_SOURCE_DIR}/openssl-1.1.0j-
 set(PATCH_CMD2 sh -c "patch -N -p1 < ${CMAKE_CURRENT_SOURCE_DIR}/openssl-1.1.1b-CVE-2019-1543.patch || true")
 set(PATCH_CMD3 sh -c "patch -N -p1 < ${CMAKE_CURRENT_SOURCE_DIR}/openssl-origin-rpath.patch || true")
 # GCC 4.9's <stdatomic.h> is broken (c.f., https://github.com/pytorch/pytorch/issues/3050)
-if($ENV{SONY_PRSTUX})
+if($ENV{SONY_PRSTUX} OR $ENV{ANDROID})


PS With your switch to OpenSSL 1.1.1 you probably just took the main (minor) obstacle to Android Clang out of the way.

[skip-ci]

NiLuJe · 2019-04-18T17:52:01Z

Crap, I added a dash to skip-ci. It's [skip ci]. -_-". IT WAS GREEN. I SWEAR! :D

Frenzie · 2019-04-18T17:53:45Z

thirdparty/openssl/CMakeLists.txt

+set(PATCH_CMD2 sh -c "patch -N -p1 < ${CMAKE_CURRENT_SOURCE_DIR}/openssl-1.1.1b-CVE-2019-1543.patch || true")
+set(PATCH_CMD3 sh -c "patch -N -p1 < ${CMAKE_CURRENT_SOURCE_DIR}/openssl-origin-rpath.patch || true")
+# GCC 4.9's <stdatomic.h> is broken (c.f., https://github.com/pytorch/pytorch/issues/3050)
+if(DEFINED ENV{SONY_PRSTUX} OR DEFINED ENV{ANDROID})


So something like ANDROID and "${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU"

That's assuming the configure script manages to pick up Clang properly (c.f. the Android branch on top with a CC=gcc to avoid it spazzing out because of ccache and our hacked CC with appended -static-libstdc++) ;o).

But, yeah, that patch would probably not behave on Clang as it's using GCC internal builtins ^^.

Frenzie · 2019-04-20T07:49:45Z

@NiLuJe There's a new error message since this commit. Not sure if it's harmless?

ccache: error: Could not find compiler "arm-linux-androideabi-gcc" in PATH

ccache: error: Could not find compiler "arm-linux-androideabi-gcc" in PATH

ccache: error: Could not find compiler "arm-linux-androideabi-gcc" in PATH

NiLuJe · 2019-04-20T13:09:09Z

At which point of the build?

…

On Sat, Apr 20, 2019, 09:49 Frans de Jonge ***@***.***> wrote: @NiLuJe <https://github.com/NiLuJe> There's a new error message since this commit. Not sure if it's harmless? ccache: error: Could not find compiler "arm-linux-androideabi-gcc" in PATH ccache: error: Could not find compiler "arm-linux-androideabi-gcc" in PATH ccache: error: Could not find compiler "arm-linux-androideabi-gcc" in PATH — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#888 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAA3KZRXPZIGYCJD4PWT763PRLDJTANCNFSM4HF3DTSQ> .

Frenzie · 2019-04-20T13:11:03Z

At every point?

Frenzie · 2019-04-20T13:11:44Z

It's clang up until the point where /usr/bin/ccache clang doesn't equal clang (due to that change), so it's switched to gcc.

Frenzie · 2019-04-20T13:13:39Z

Oh, you meant the ccache "could not find in path" error message. Pardon. Um, that's somewhere near the beginning.

Frenzie · 2019-04-20T13:15:22Z

Anyway, the clang build is now broken because that /usr/bin/ccache clang mismatch made it fly under the radar.

cc1plus: warning: unrecognized command line option "-Wno-macro-redefined"
cc1plus: warning: unrecognized command line option "-Wno-parentheses-equality"
cc1plus: warning: unrecognized command line option "-Wno-unused-command-line-argument"
cc1plus: warning: unrecognized command line option "-Wno-unknown-warning-option"
cc1plus: warning: unrecognized command line option "-Wno-ignored-optimization-argument"

Frenzie · 2019-04-20T13:17:12Z

Never mind(ish), it seems to be using GCC there for some reason. >_< So the Clang build is still broken, just not due to Clang per se.

A bit of an emergency fix to just get it working ASAP. A patch was added to base LuaJIT in koreader/koreader-base#892, which caused a mismatch and build failures over here. Unfortunately that was hidden by some excessive Android verbosity inadvertently introduced in koreader/koreader-base#888, and on my local instance where I did a sanity check LuaJIT was already compiled…

* [fix] Add LuaJIT patch to sync with koreader-base koreader/android-luajit-launcher#141 A bit of an emergency fix to just get it working ASAP. A patch was added to base LuaJIT in koreader/koreader-base#892, which caused a mismatch and build failures over here. Unfortunately that was hidden by some excessive Android verbosity inadvertently introduced in koreader/koreader-base#888, and on my local instance where I did a sanity check LuaJIT was already compiled…

NiLuJe mentioned this pull request Apr 14, 2019

Allow toggling CRe's new dithering & scaling koreader/koreader#4922

Merged

poire-z approved these changes Apr 14, 2019

View reviewed changes

Frenzie reviewed Apr 15, 2019

View reviewed changes

NiLuJe force-pushed the cre-img branch from 2152ce7 to 1eac88b Compare April 17, 2019 15:14

NiLuJe added 25 commits April 18, 2019 00:26

Build the new scaler code

4b9f088

Allow toggling smooth scaling

4f9861d

Now with the right function name ;D

32425f3

Lay the groundwork to make SW/HW dithering easily toggleable

1bacd69

Allow toggling SW dithering in CRe

d69ced1

Don't even try to enable SW dithering when !8bpp, as nothing will honor

9e87b55

it anyway

Initial naive port of the dithering algo to the Lua BB

7aa3810

Simplify that

345b073

And port dithering to the C BB

9cd67b9

And now it actually builds ;p

e41ca23

Ups.

3e9cc31

Nice typo.

c265656

Make sure we pass a Color8 to the dither algo in the Lua BB

9def095

Declare dither_o8x8 earlier for ditherpmulblend

41135d6

Jot that down for one rainy day of experimentation...

babdd9f

Bump CRe to pickup the required dithering/scaling changes

a4cbade

Let's see PB's GCC choke on C11/C++11...

Heh. That's what happens when DIV255 gets more accurate ;).

ed5e6d9

Enforce C11/C++11 w/ GNU extensions on older compilers.

c394ecf

Here's hoping that'll do the trick even on PB...

And explain why GCC 6

ffb8c5c

(IIRC, GCC 6.0 never came out, so, eh).

More comments :)

59f4e17

Bypass ccache when version checking (maybe?)

dc86eff

Vague attempt at getting CMake to actually use the right flags

45f8cbf

It's calling CXX w/ CFLAGS, which is braindead on all kinds of levels.

Oh, CMake, how I love how you pull bogus CFLAGS out of nowhere.

9bb7c33

Nope, CMake is still the worst.

0433298

Slightly smaller hammer should be enough.

c33d180

NiLuJe added 11 commits April 18, 2019 01:07

Err, set API level on both Android variants -_-".

a328007

Well, the soft approach didn't work, I see...

62ee45c

Ups, we install a standalone toolchain...

e7749e4

Keep those around, just in case...

db57c8f

Third time's the charm?

21da5b8

Okay, don't override CC on Android

aa775de

Or do, because ccache is awful.

08b7b68

That was stupid ;)

a88a5f4

I was expecting that one ;).

fe59473

Anbdroid TCs are stuck on GCC 4.9 (Pre-GPL3).

*sigh*

8a7cad4

The tests don't build on Android, and we don't care anyway.

accf876

test/rsa_complex.c:17:23: fatal error: complex.h: No such file or directory

Frenzie reviewed Apr 18, 2019

View reviewed changes

Fix PHONY targets

22dac0a

[skip-ci]

Frenzie reviewed Apr 18, 2019

View reviewed changes

NiLuJe merged commit b6b6d74 into koreader:master Apr 18, 2019

Frenzie mentioned this pull request Apr 22, 2019

[fix] Add LuaJIT patch to sync with koreader-base koreader/android-luajit-launcher#141

Merged

Frenzie mentioned this pull request Apr 22, 2019

[Android] Bump luajit-launcher to fix build koreader/koreader#4957

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Allow toggling CRe's new dithering & scaling #888

Allow toggling CRe's new dithering & scaling #888

NiLuJe commented Apr 14, 2019

NiLuJe commented Apr 15, 2019

Frenzie Apr 15, 2019

NiLuJe Apr 15, 2019 •

edited

Loading

Frenzie Apr 18, 2019

NiLuJe Apr 18, 2019

NiLuJe commented Apr 18, 2019 •

edited

Loading

Frenzie Apr 18, 2019

NiLuJe Apr 18, 2019 •

edited

Loading

NiLuJe Apr 18, 2019 •

edited

Loading

Frenzie commented Apr 20, 2019

NiLuJe commented Apr 20, 2019 via email

Frenzie commented Apr 20, 2019

Frenzie commented Apr 20, 2019

Frenzie commented Apr 20, 2019

Frenzie commented Apr 20, 2019

Frenzie commented Apr 20, 2019 •

edited

Loading

Allow toggling CRe's new dithering & scaling #888

Allow toggling CRe's new dithering & scaling #888

Conversation

NiLuJe commented Apr 14, 2019

NiLuJe commented Apr 15, 2019

Frenzie Apr 15, 2019

Choose a reason for hiding this comment

NiLuJe Apr 15, 2019 • edited Loading

Choose a reason for hiding this comment

Frenzie Apr 18, 2019

Choose a reason for hiding this comment

NiLuJe Apr 18, 2019

Choose a reason for hiding this comment

NiLuJe commented Apr 18, 2019 • edited Loading

Frenzie Apr 18, 2019

Choose a reason for hiding this comment

NiLuJe Apr 18, 2019 • edited Loading

Choose a reason for hiding this comment

NiLuJe Apr 18, 2019 • edited Loading

Choose a reason for hiding this comment

Frenzie commented Apr 20, 2019

NiLuJe commented Apr 20, 2019 via email

Frenzie commented Apr 20, 2019

Frenzie commented Apr 20, 2019

Frenzie commented Apr 20, 2019

Frenzie commented Apr 20, 2019

Frenzie commented Apr 20, 2019 • edited Loading

NiLuJe Apr 15, 2019 •

edited

Loading

NiLuJe commented Apr 18, 2019 •

edited

Loading

NiLuJe Apr 18, 2019 •

edited

Loading

NiLuJe Apr 18, 2019 •

edited

Loading

Frenzie commented Apr 20, 2019 •

edited

Loading