Work around Arm Compiler 6 stdvector perfomance cliff

[theotherjimmy]

@sg-

[theotherjimmy]

Size diff

fat-fs example --profile release

Before

Text 59.2KB Data 224B BSS 9.35KB ROM 59.4KB RAM 9.56KB

After

Text 59.2KB Data 224B BSS 9.35KB ROM 59.4KB RAM 9.56KB

fat-fs example --profile develop

Before

Text 62KB Data 216B BSS 9.35KB ROM 62.2KB RAM 9.56KB

After

Text 62KB Data 216B BSS 9.35KB ROM 62.2KB RAM 9.56KB

[theotherjimmy]

Note: The commits are backwards in github.

More Size diff

tests-mbed_drivers-stl_features, --profile release

9145898

| Subtotals                   | 134952 |   280 | 13146 |

2fd7cfc

| Subtotals                   | 55182 |   280 | 9679 |

[0xc0170]

@theotherjimmy Can you clarify (provide more details in the commit message what is this fixing and how) ?

The sizediff in #5993 (comment) are the same.

[sg-]

LGTM 👍

Can we also get numbers for the same test using ARMC5?

[theotherjimmy]

Can we also get numbers for the same test using ARMC5?

The target branch does not support ARMC5. I'd have to retarget, or open a different PR.

[pan-]

@theotherjimmy That's better 👍 .

Hopefully once the version of libc++ shipped with ARMCC will be released, we will be able to use the configuration macro _LIBCPP_DISABLE_EXTERN_TEMPLATE.

[0xc0170]

/morph build

[mbed-ci]

Build : FAILURE

Build number : 1107
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/5993/

[pan-]

The error is comming from ble; I've proposed a fix in #6058 .

[cmonr]

Aaand it's rebase time!

[0xc0170]

/morph build

[mbed-ci]

Build : FAILURE

Build number : 1173
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/5993/

[0xc0170]

@theotherjimmy If you can rebase the destination branch to include the fix ?

[theotherjimmy]

@0xc0170 I'll rebase now

[theotherjimmy]

/morph build

[theotherjimmy]

Rebase done for both feature-armc6 and this PR.

[mbed-ci]

Build : FAILURE

Build number : 1183
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/5993/

[cmonr]

Logs look like they were suddenly cut off. Rebuilding.

/morph build

[mbed-ci]

Build : FAILURE

Build number : 1187
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/5993/

[theotherjimmy]

This should go to master

[cmonr]

@theotherjimmy Like so?

[0xc0170]

Destination branch updated, seems like some changes should not be here, please review and rebase

[theotherjimmy]

@0xc0170 @cmonr Rebased.

[cmonr]

/morph build

[mbed-ci]

Build : SUCCESS

Build number : 1306
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/5993/

Triggering tests

/morph test
/morph uvisor-test
/morph export-build
/morph mbed2-build

[mbed-ci]

Exporter Build : SUCCESS

Build number : 966
Build artifacts/logs : http://mbed-os.s3-website-eu-west-1.amazonaws.com/?prefix=builds/exporter/5993/

[mbed-ci]

Test : SUCCESS

Build number : 1090
Test logs :http://mbed-os-logs.s3-website-us-west-1.amazonaws.com/?prefix=logs/5993/1090

[michalpasztamobica]

Hi @pan- and @theotherjimmy , I'm wondering if the -D_LIBCPP_EXTERN_TEMPLATE(...)= flag can be removed now?
@pan- mentioned this would be the case if ARMCC was shipped with libc++. I haven't found a particular changelog but it seems to me that it is shipped with llvm's libc++.
The flag increases the binary size and generates warnings, when std:: is being put to use.
I can set up a PR, but before I do, I thought I'd better just ask in here, in case I am missing something important.

[theotherjimmy]

@michalpasztamobica You may want to open a new issue to discuss the removal of this workaround. Further, Note that I no longer work on Mbed OS, and I would appreciate being left out of these discussions. @ARMmbed/mbed-os-tools Has taken over my role.