Maintainers of open source projects should be mentoring folks to become their replacements. Not just thinking this is about controlling their project.
I have learned this the hard way for gcc.
You can't just think it is your way or the highway. Also gsoc is NOT a replacement for mentoring inside the project. It should supplement the mentoring that is done, just like an internship.

Also this is not unique to gcc, it is many open source projects.
And then there is the gaming of github stats, the one or two commits in llvm and leave issue. Those are used to gain stats for a job.
Seriously open source is broken because of the gaming of the system.

Even inside companies where you have the person with the most open source commits. Or the person who opened the most issue. Or the person who reviewed the most. (What is funny is that i am that person but I am not trying to play the game of counting, I am just trying to push gcc forward).
At the same time I am mentoring folks both inside and outside of qualcomm on both gcc (and even to a less sense llvm) and performance analysis.
But going back to my original point is mentoring and maintainship overlaps. It is not one or the other but rather both in the same role.

By the time you step down you should have at least mentored a few to be able to take you from you. If you have not you partially failed a as a maintainer as the project will no longer can be maintained in the future.

16 years ago today my wife and I got married. We have got through a lot in these last 16 years including moving to a new state, 8 years ago; me going to China for work for months at a time (she still jokes about that when I go on a trip to a conference; she says I am going to China). Me changing jobs 3 years ago. Losing my wife's dog (that is still hits the hardest for my wife) 5 years ago and then losing her two cats within a year of this. Getting two new cats and a new dog. I love my wife so much.

How did Ubuntu become the most popular Linux distro?
Was it just because it was free and looked to be the most professional one?

Centos was the most popular for most businesses until Ubuntu came about.
And the funny part is canonical always been going out of business. So why do so many depend on a distro which is dependent on a company that is going out of business for the last 15 years?

RE: https://hachyderm.io/@dalias/116625664185661900

Most new open source projects have gone the whole idea direct to users is the way to go.

This has caused the idea that you can only support a few things and that then CI that one thing to death. Rather than the idea distros are out there to help spread the burden of testing/support and have a layer to help all spread the support to more than just the developers.

This is why for an example GCC does NOT do binary releases at all. It would be burden to directly support those even on a select few distros (which ones again? CentOS 8, 9 or 10? Ubuntu 2404 or the latest LTS? SuSE xx? what about gentoo, etc.).
So distros help spread the support , testing layers and even come back to the developers saying this is broken and many of the times how to fix it.
Developers only need to do one testing but distros help both sides. Even help by doing development and even push for a new feature (to the extend not being pushy maybe even help by getting funding directly from the users to develop it).

I hacked on some python code today and the error messages from python was just so bad.
Why does people like python again?
I forgot to closing quote and it was not obvious from the error message I forgot it.
A simple thing as that gives a
Syntaxerror EOL while scanning string literal.

What the hell does that mean? Oh right I forgot a end quote. Not even pointing to the beginning of the string which would have made it obvious. Nope just a syntax error message that seems like from the 80s.

Also "engage with WebAssembly and LLVM projects." Is funny because "safety C++ and C" are LLVM projects that Apple is leading. This person just does not know what they are talking about.

RE: https://mastodon.social/@alexa_pavlova/116453973860406617

The bug was not really a bug. It was a loophole that Apple didn't realize until it was too late.
Also it was a logic issue. Rust would NOT fix that issue.
Folks like I quoted here is exactly why I mention rust is a cult.
Also Apple is moving towards safety C++ and C faster than most other company because they control programming langauge inputs. Unlike most other places like Linux (open source). It is both a blessing and a problem.

@regehr might know the answer to this. Are there any papers/programs about fuzzing compilers with C++ code?

I have not seen anyone use one with GCC (nor LLVM bug reports) so far and the main reason why I ask is because of I have now seen some bugs dealing with (speculatively) devirtualization in GCC.

Let's talk about why compilers and static analysis for a second.

Most companies are paying folks to work on compilers to improve code generation and not looking into other areas (security).
It is only application developers who are pushing for better tooling in those same companies. But that priority message does not get down to the compiler team.

And then when compiler folks implement something (e.g. -Wstringop-overflow, Warray-bounds), application developers complain about them because they were not tested on their applications and provide too many false positives.

Oh did I mention there are also some closed source static analyzers out there which companies are just paying money for. Instead of investing in open source world.

So when things like LLMs come around and can do "static analyze" for cheaper and seemly maybe better than the closed source programs well companies are going to jump on it.

Basically what I am pointing out is the problems with current state of things when it comes to static analyzers.
I don't know anyone who even tries to use clang analyzer and I have not even seen anyone prompting it. On the GCC side it does seem like there is a little more prompting of GCC analyzer and getting open source developers to try it out and report issues and even GSOC projects to improve it. But GCC's analyzer is much newer and does not cover a lot.