#timvideos

shenki: I think you have cut off the error there

shenki: Timing: Completed - 82 errors found.

mithro: why did i get different results to travis?

shenki: I can't see the errors, so I can't see why it is different?

mithro: okay. which log file do we want? i've closed the tab so i don't have the console history

shenki: a couple of pages above that output would be useful

shenki: as well, ISE gets confused sometimes

ok, i'll rebuild

mithro: should i delete anything before rebuilding?

shenki: no need

ok

and this is what i want to type?

make BOARD=atlys TARGET=hdmi2usb PROG=openocd gateware

no

you shouldn't need anything more than "make gateware" all the others are defaults

k

mithro: it worked this time :/

mithro: thanks for your help ")

:)

shenki: good to see you trying things :P

[mithro/HDMI2USB-misoc-firmware/master#55] (480cd42): The build passed. (https://travis-ci.org/mithro/HDMI2USB-misoc-fir...)

mithro: thanks :) now that i've got a hdmi2usb and a hdmi2eth build, i will test it tonight

i want to demo the hdmi2eth to a friend

mithro: can you help me with a c++ question?

mithro: you were going on about const expressions the other day

shenki: I'm around know if you still need that C++ help

mithro: i have this snippet of code

static Target* const MASTER_PROCESSOR_CHIP_TARGET_SENTINEL

= (sizeof(void*) == 4) ?

reinterpret_cast<TARGETING::Target* const>(0xFFFFFFFF)

: reinterpret_cast<TARGETING::Target* const>(0xFFFFFFFFFFFFFFFFULL);

it's in a header

and i want it to be a compile time constant

gcc warns me that it's unused, which i thought it wasn't supposed to do for static const things

(i didn't write the code. im trying to build it using a newer compiler)

shenki: what C++ version are you compiling with?

shenki: that expression isn't a valid const expression in older C++ versions IIRC

shenki: Until C++11 things you think should be constant expressions are not

mithro: im not sure which version. let me examine the flags

mithro: you suspect that if i do -std=c++11 it might work?

shenki: It'll be closer to working :)

heh ok

Title: Constant expressions - cppreference.com (at en.cppreference.com)

mithro: the warning im getting is relating to an unused var

mithro: i thought i wanted to fix it by making the whole the const

coz normally in c++ you can have a static variable instead of doing #define?

shenki: btw is the pointer const or is it pointing to something that is const, or both?

can you past the whole error

?

it should be both

it's not in the code as it currently exists

I think you need more const in there :P

yeah, that might be the issue

then all of the call sites blow up :/

but yeah, i think the issue is a long standing bug

and gcc 4.9 has a bug where it doesn't produce any unused warning for it

In file included from mdiafwd.H:34:0,

from mdiamba.C:30:

../../../../src/include/usr/targeting/common/targetservice.H:137:22: error: ‘TARGETING::MASTER_PROCESSOR_CHIP_TARGET_SENTINEL’ defined but not used [-Werror=unused-variable]

static Target* const MASTER_PROCESSOR_CHIP_TARGET_SENTINEL

^

cc1plus: all warnings being treated as errors

shenki: and you sure you are using it somewhere?

well, yes and no. it's a header, so sometimes it's used and sometimes it's not

depends on the c++ file

but C++ is not supposed to warn about stuff that's static const, right?

shenki: I don't see why it wouldn't?

$ g++ -Wall -O2 -c a.cc

$ echo $?

0

$ cat a.cc

static const int a = 10;

coz c++ developers use static const instead of #defines

i dunno. perhaps im all confused

shenki: but if you never use your #define anywhere?

btw you shouldn't be putting that initializer in the header file

ok. why's that?

(i need ammo to tell the peeps who wrote it why they are wrong)

shenki: it's equivalent to putting a function body in the header

only if it's not const, right?

static char const b = 1 ? 'a' : 'b';

that line doesn't emit a warning

static char* const b = 1 ? reinterpret_cast<char* const>('a') : reinterpret_cast<char* const>('b');

neither does that

but if i change the 1 to sizeof(void *) it does warn

so that means that sizeof(void *) is not constant?

which is bullshit, the compiler should know that at compile time

I think the compiler doesn't see "sizeof(void*)" as a valid constexpr

so it ends up actually allocating the value

yeah

i agree

i proposed they use __SIZEOF_POINTER__, a builtin define that gcc provides

shenki: if you look at a objdump of the output elf file, you'll see it

good idea

what does "sizeof(void const*)" ? or actually = "sizeof(TARGETTING::Target* const)" do?

shenki: IE its weird to do the sizeof to a different thing you are reinterp casting from?

it's a sentinel value

so they're trying to make a pointer whos vlaue is -1

ie, all 1s

0xFFF....etc

the sizeof is to find out of a pointer should be 64 1s, or 32 1s

shenki: sure - I understand that

right

you're saying why not sizeof(Target *) ?

8:16 PM <+shenki> = (sizeof(void*) == 4) ? reinterpret_cast<TARGETING::Target* const>(0xFFFFFFFF) : reinterpret_cast<TARGETING::Target* const>(0xFFFFFFFFFFFFFFFFULL);

i guess it's a pointer, it doesn't matter what type of pointer you sizeof

= (AAAA == 4) ? reinterpret_cast<AAAA>(0xFFFFFFFF) : reinterpret_cast<AAAA>(0xFFFFFFFFFFFFFFFFULL);

Logically, all the AAAA should be the same I think

yeah. that's good style

but sizeof(<anything> *) will always be the same value

so functionally there's no difference

it could do sizeof(int *) for all it matters

this code is just silly

i give up

shenki: depends on your architecture :P - on sdcc a sizeof(int*) != sizeof(int*) depending on which memory the int* is in

heh, okay

mithro: got the firmware loaded onto my atlys

shenki: great, didn't we do that while I was at your place?

yeah. but now ive done it by myself

im trying to work out how the ethernet works. it looks like there's nothing on the lm32 for it

shenki: yeah, I don't know very much about it

ok

mithro: what's our max res/framerate on usb these days?

shenki: I'd start with seeing if the link light comes up / pinging it / etc

shenki: 720p30

yeah, i get the link light

okay, thanks

shenki: you should actually be able to set it up so that you can tftp boot the lm32 firmware

mithro: i did see that. i'm poking around

the ip is the same range as my home network