Binary size in daily report is same for all targets about js-remote-test HOT 10 CLOSED

@glistening: nuttx.bin contains the tests (by ROMFS) as well. Should we have a NuttX build without the tests and report that size?

from js-remote-test.

@rtakacs Yes. We would like to know the minimal size. (i.e. without debug symbol, unit tests, ...)

from js-remote-test.

One way to measure it is to use arm-none-eabi-size to get the size from static library file(*.a).

from js-remote-test.

In the build/arm-nuttx/release/lib folder there are more static libraries:

libhttpparser.a libiotjs.a libjerry-core.a libjerry-ext.a libjerry-libm.a libtuv.a

Is it enough just aggregate the main segments (.text, .data, .rodata) to approximate the binary size? In this case the binary size of the RPi2 target also should be use this mechanism (and not the ls command that currently is used).

The binary size chart on the page can be an aggregated chart that can show the .text .data .rodata values.

from js-remote-test.

@rtakacs It seems ok. But I would like to get confirmed from @hs0225.

from js-remote-test.

The archive files contain unused symbols too. Link time optimization drop those symbols from the final binary, so reporting the archive sizes could be misleading. What about analyzing the map file of the linker? It might give us a more precise result.

from js-remote-test.

@LaszloLango Yes, you are right. Maybe the map file should be processed, because that contains the final .text, .data, .rodata sizes.

from js-remote-test.

@LaszloLango, @rtakacs In general, you're right. But @hs0225 suggested this approach for some reason. He showed numbers that it shows not much differences. He will contact and explain you on IRC for details. I am ok for map approach if @hs0225 agrees.

from js-remote-test.

@LaszloLango @rtakacs I agree to use the map file because it's the right way. FYI, TizenRT provides a utility to analyze map files.(os/tools/memstats_gnueabil.py)
Is it possible to measure size by using map file in rpi2 and nuttx?

from js-remote-test.

@hs0225 Yes, #53 uses the linker map file (on both targets) to approximate the binary size. Since #53 is landed, I'm closing this issue.

from js-remote-test.