Make target provide jerry-port.c

[franc0is]

This change-set pushes ports forward a bit and adds a Pebble port (happy to remove the latter if it is inappropriate).

JerryScript-DCO-1.0-Signed-off-by: François Baldassari francois@pebble.com

[zherczeg]

Please check CI, and I recommend to run make precommit before uploading a patch.

Signed-off-by message is incorrect. The following line should be at the end of the 3413e96 commit's message: 'JerryScript-DCO-1.0-Signed-off-by: François Baldassari francois@pebble.com'.

[akosthekiss]

As you also mention in your comment, this patch does several things. Since some of them are easier to accept than others, I'd suggest splitting it up to several PRs.

• I'm all in for the use of jerry_port_errormsg in jrt-fatals. However, there are macros in jrt.h to use it (JERRY_ERROR_MSG).
• The use of jerry_port_putchar in ecma-builtin-global.c is incorrect. E.g., current implementation prints the char literals backslash-u-zero-zero-zero-zero for a NUL char, while the suggested change here would print the NUL char itself. Similar incompatible change happens to the two-byte code points. It can be OK to use the port function to output a char to the console instead of printf but then it should be done correctly.
• About extending the port API, I haven't come to a consensus with myself yet. I have a feeling that it starts duplicating the libc API with slightly different names, which is somewhat disturbing.

[franc0is]

@zherczeg I am planning on squashing that commit after I address other comments. Thanks for taking a look!

[franc0is]

@akiss77 Thank you for the thoughtful comments. I'll try to address them as best I can:

As you also mention in your comment, this patch does several things. Since some of them are easier to accept than others, I'd suggest splitting it up to several PRs.

OK - I think I can split it into 2 to start. One for the port, one for some of the CMakelist changes

I'm all in for the use of jerry_port_errormsg in jrt-fatals. However, there are macros in jrt.h to use it (JERRY_ERROR_MSG).

Ah, I did not see this. Will change.

The use of jerry_port_putchar in ecma-builtin-global.c is incorrect. E.g., current implementation prints the char literals backslash-u-zero-zero-zero-zero for a NUL char, while the suggested change here would print the NUL char itself.

As best I can tell, those two are equivalent. \u0000 is the unicode representation of the NUL character.

Similar incompatible change happens to the two-byte code points. It can be OK to use the port function to output a char to the console instead of printf but then it should be done correctly.

You are right. I need to encode those bytes before I output them with putchar. Thanks for catching this, I will fix it.

About extending the port API, I haven't come to a consensus with myself yet. I have a feeling that it starts duplicating the libc API with slightly different names, which is somewhat disturbing.

The reason I introduced jerry_port_abort rather than simply rely on libc functionality is that on my platform I want to do something specific to jerryscript (i.e. clean-up the JS context and show a dialog on the device) whereas libc's abort is necessarily more broad.

[franc0is]

Split #955, #956, and #957 out of this PR

[LaszloLango]

Please use a better name

[franc0is]

This is the name of the abort API internal to the Pebble firmware, I cannot change it easily. Only option is to remove the Pebble port from this PR.

[franc0is]

I removed the pebble port (it is of little use to those without access to the pebble firmware anyways), so this PR is quite simple now.

[franc0is]

Ping! I think this does not overlap with the proposal in #964 and should be considered individually.

[LaszloLango]

LGTM

[zherczeg]

LGTM

[franc0is]

rebased.