On Mon, Mar 06, 2023 at 04:15:59PM -0500, Jeremy Bícha wrote: > Michael Catanzaro advised us not to use -g1 to build webkit2gtk at > least on amd64 to get better crash reports.
I was talking about this with Michael some days ago. As you said, he thinks that we should use -g in the Debian packages. Here's how I see it: - Stack traces with -g1 don't show the values of the parameters and are not really useful for debugging WebKit with gdb. - Still they show function names and line numbers, so we can see where the crash happened and in many cases that's enough information to identify the problem and work on it. - With -g1 each one of the main -dbgsym packages takes almost ~160MB of disk space, whereas with -g it's closer to 1GB per package. That's a very significant difference and I fear that it might be enough to discourage some users from installing the -dbgsym packages when there's a crash (I seem to remember that it actually happened once or twice). - Builds with -g are also more cumbersome to handle, requiring a significant amount of disk space to build. - WebKitGTK is also a project with frequent releases, and that means several GBs worth of new binary packages every few weeks. - In all these years I received many bug reports, some of them with stack traces. I don't remember a single instance where a stack trace produced with -g would had made a difference. Michael also admitted that this is not a common situation in his experience. For all these reasons, and although I think that building with -g is the right thing to do if you are developing WebKit, for end users I think that using -g1 is a reasonable trade-off. I don't know if you see things differently? I'm open to reconsider. Berto
