The last day ended with the Linux kernel developer panel discussion, Linux Kernel Developer Panel.
The moderator is GREG KROAH-HARTMAN (The Linux Foundation, hereafter GREG), a maintainer of the Stable Kernel.As a panelist, THOMAS GLEIXNER (Linutronix), a maintainer such as real -time patches, Mr. Zefan Li (Huawei), one of Cgroup's maintainers, and a bridge, Stephen Hemminger (VYATTA), a maintainer in the network field.Grant Likely (LINARO), a maintainer of Device Tree, and four different backgrounds.
モデレーターのGreg Kroah-Hartman氏(The Linux Foundation)左から、Thomas Gleixner氏(linutronix)、Zefan Li氏(Huawei)、Stephen Hemminger氏(Vyatta)、Grant Likely氏(Linaro)In the first self -introduction, GREG gave the title "What is the first patch?"GREG's first patch was related to USB.
On the stage, discussions on Device Tree were held.Likely explains Device Tree and the technology that appeared in the embedded field.It is said that the kernel can refer to the hardware specification data so that the kernel can be referenced, and it has been adopted as a solution to various hardware differences.On the other hand, the ARM server has adopted ACPI, and the ARM kernel has also added ACPI support.Likely argued that the two look the same but different, and that Device Tree will not disappear for the time being.
In addition, there are dissatisfaction with the change of Cgruop's interface, and that discussions have occurred regarding the expansion of Netlink in the network field, and "Regarding the change, even if it goes well in that part, it is overall.It is difficult to judge, "If you publish the API, it will be difficult to change later," "It will take many years to judge", "It will be necessary for those who have experience in the interface to be considered.There was an opinion.
A question was actively asked from the venue.Regarding the code test, you can find a bug, so you can help, you need a test when you change it, but you have a voice that you do not perform regularly, and if it is broken by preparing an automatic build tool, you will receive an email.There were stories about the fact that the combination of build configuration was difficult.
Regarding the question of how Menteners responded to large amounts of emails and efficiently reviewing patches, GREG first answered that they were using command line email software to use scripts.When Hemminger answered, "The priority varies depending on the type of patch," Gleixner said, "In addition, patches for those who have a track record in the past are reliable, but those who have put out terrible patches in the past.In some cases, it is necessary to take a long time to review it. "Likely said, "I learned that the patches that I trusted by the person I trust do not need to review."Li also said, "I was in trouble because I didn't know the patch when the problem occurred in Cgroup and MySQL in a non -familiar field."
Regarding the question "Do you use Stable Kernel on a daily basis", GREG, a maintainer of Stable Kernel, answered, "I don't use it."Panelists answered a variety of answers, including those who use the latest kernel and those who use Stable Kernel to use Stable Kernel on a virtual machine.In addition, the question "Do you hack your smartphone kernel?" From the answer that you used to do it in the past, but you haven't done it now, and the answer that you haven't done it at all, and the answer "I don't have a smartphone (laughs)".rice field.
When asked about English issues, it developed from the voice "Changelog (a change history document)", which gave advice on Changelog.Panelists commented that it is important to write "why so" instead of "how they changed" one after another.This is because the intention is not understood from the code, it is for explanation of why the function is necessary, and if you do not approach the purpose during the review, you can advice, and I saw it two years later.It was explained that sometimes I didn't know why.
The question "What I want to convey to a new kernel developer" was told an enthusiastic advice.GREG advised, "Let's do it first because I'm interested."Gleixner said, "I don't have to be disgusted. Even if the patch code is wrong, if you point out the problem, it will be a guide for other people."When submitted, the reviewers are tired, so I want you to read the feedback carefully and think about it. "
Li advised, "Looking at the past patches, looking at the code that has made a major change, and learning how to change it.""I want to find a mentor. I'm trying to say a positive word," said Hemminger in a way.Likely said, "I didn't know what to do at first, but when I joined Linux Symposium and met the developer, I knew my opponent's face and became easier to send patches."