#0.11.0 Windows backtrace not symbolizing

1 messages · Page 1 of 1 (latest)

whole junco
#

Hey zigfriends, 0.11.0 windows build seems to not symbolize backtraces properly out of the box. any std lib hacks I can do to make it work?

My setup is:

running in vscode via a cppvsdebug debug launch config using whatever zig is in 5PATH%, zig build is 0.11.0 installed via zigup

I've seen partial symbolication success with catch unreachable, but it seemed to die in recursive functions.

What I'm seeing now is GPA's leak report is just listing addresses like so:

???:?:?: 0x7ff66a3bd0cd in ??? (???)
???:?:?: 0x7ff66a381d7a in ??? (???)
???:?:?: 0x7ff66a381bee in ??? (???)
???:?:?: 0x7ff66a37f7df in ??? (???)
???:?:?: 0x7ff66a37f0e7 in ??? (???)
???:?:?: 0x7ff66a520d4d in ??? (???)
???:?:?: 0x7ff66a5167a1 in ??? (???)
???:?:?: 0x7ff66a50a206 in ??? (???)

error(gpa): memory address 0x1cf4aba0040 leaked: 
???:?:?: 0x7ff66a3bd0cd in ??? (???)
???:?:?: 0x7ff66a381d7a in ??? (???)
???:?:?: 0x7ff66a381bee in ??? (???)
???:?:?: 0x7ff66a37f7df in ??? (???)
???:?:?: 0x7ff66a37f0e7 in ??? (???)
???:?:?: 0x7ff66a520d4d in ??? (???)
???:?:?: 0x7ff66a5167a1 in ??? (???)
???:?:?: 0x7ff66a50a206 in ??? (???)```

(truncated, I have a bunch of leaks 😛 )
#

0.11.0 Windows backtrace not symbolizing

whole junco
#

I compiled & ran my code on nightly and backtraces symbolicated properly
I can temporarily use master to fix my leaks, but if somebody knows of a local change to fix this I'm eager to know. happy to maintain a local patch if it's a simple std lib thing

whole junco
#

Anyone know of a workaround for 0.11.0?

merry pendant
#

Why would you would rather use a patched 0.11.0 than a nightly build?

whole junco
#

I did nightly for the 0.10.x era and usually would have to maintain local patches anyway for temp workarounds, so maintaining that as fixes got merged was more annoying than if I'd just patched 0.10.0

also seems like zls doesn't have auto fix for the new var/const errors that are on nightly, so I wanna avoid that for a while

#

also, in the past the fixes for this stuff has just been easy std lib edits

blazing ravine
#

also seems like zls doesn't have auto fix for the new var/const errors that are on nightly, so I wanna avoid that for a while
sure it does, on the master branch

whole junco
#

oh that's good to know, the open issue and discussion from yesterday had me thinking folks were still working on it

blazing ravine
#

we sync these sorts of things

#

zls was fixed immediately

whole junco
#

hell yeah. I'll try pinning to a recent nightly zig & zls then
thanks for confirming 🙂

whole junco
#

oh when you say master, you mean an actual fresh build of master, not just the latest binaries?

blazing ravine
#

depends for what

whole junco
#

getting a build of zls with the autofix for unmutated var

blazing ravine
#

yeah the head

#

you need to use the head