Bug 5478

Summary:

test build time performance degradations with gcc security flags enabled

Product:

[QA/Testing] Test Plans/Suite

Reporter:

Alexandru Georgescu <alexandru.c.georgescu>

Component:

test-plans/suite

Assignee:

Humberto Ibarra <humberto.ibarra.lopez>

Status:

RESOLVED OBSOLETE

QA Contact:

Severity:

enhancement

Priority:

Medium+

CC:

alexandru.c.georgescu, bogdanx.a.voiculescu, corneliux.stoicescu, jessica.zhang, jose.perez.carranza, sgw

Version:

1.6

Target Milestone:

2.3 M2

Hardware:

x86

OS:

Multiple

Whiteboard:

OS type for building Yocto:

---

Type of Regression:

---

Verified:

Documentation change:

Don't know

Bug Depends on:

Bug Blocks:

4144, 5477

Attachments:

Description	Flags
GCC security flags - build time tests	none

Description Alexandru Georgescu 2013-11-05 16:35:31 UTC

Sub bug for 5477. We need to test how build performance is affected with and without gcc security flags (see bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=3868 for details).

We will test the following images:
core-image-minimal, core-image-sato, core-image-sato-sdk.

The build times will be presented and compared to see the build time affected.

The test should be done with downloads already ready, hence the download time won't be affected.

Other comments are welcomed.

Comment 1 Alexandru Georgescu 2013-11-07 15:44:23 UTC

We should use buildhistory to track the performance too.

Comment 2 Alexandru Georgescu 2013-11-08 10:52:22 UTC

we need to check build size as well

Comment 3 Ionut Chisanovici 2013-11-11 10:15:29 UTC

Created attachment 1612 [details]
GCC security flags - build time tests

Comment 4 Ionut Chisanovici 2013-11-11 11:09:08 UTC

Hi,

As we can see from the attached doc the differences within the build times are as follows:
*core-image-minimal: around 20 seconds more for the image with the security flags enabled
*core-image-sato: around 1 minute more for the image with the security flags enabled
*core-image-sato-sdk: around 2 minutes more for the image with the security flags enabled

Also, you can see that the occupied disk space differences between the two core-image-sato-sdk images are around 1 GB more for the images having security flags enable.

Having the above values, in my opinion, we can consider them to be negligible as these differences can appear on any normal build time.
Speaking only from the build time perspective, the security flags have a minimum impact on this process.

Comment 5 Alexandru Georgescu 2014-04-23 09:04:51 UTC

closing wince [qa] tracking bugs won't be used starting 1.7

Comment 6 Alexandru Georgescu 2014-04-24 09:46:17 UTC

reopened since this is not a tracking bug.

Comment 7 Ionut Chisanovici 2014-07-08 09:09:55 UTC

Moved to M4 as we want to use Lava server as a data collection tool for the performance testing.
As agreed with RP and Paul E., we will first start with Lava installation then we should start collection boot time for the images.

Cheers,
Ionut C

Comment 8 Alexandru Georgescu 2015-11-10 11:24:59 UTC

Added an estimate but this needs to be re-discussed.

Comment 9 Humberto Ibarra 2017-01-22 13:10:09 UTC

Not sure why this was blocking the runtime performance umbrella bug since it is build time testing, either way, the whole structure for those bugs is being changed. These bugs were 3-4 years old and might not be reflecting the current needs of this framework. This is being reevaluated and if it is still needed a new bug will be opened attached to the new umbrella bug #10951. It will also be reviewed if it needs to be this specific or if a broader definition for the bug is better.