update rust-minidump to e954907c
Categories
(Socorro :: Processor, task, P2)
Tracking
(Not tracked)
People
(Reporter: willkg, Assigned: willkg)
References
Details
Attachments
(2 files)
We should update to rust-minidump to pick up the fix for bug #1757890. 9cd4c9fb corresponds with 0.10.0 release.
|
Assignee | |
Updated•3 years ago
|
|
|
Assignee | |
Comment 1•3 years ago
|
||
We also need to switch github users:
|
|
Assignee | |
Comment 2•3 years ago
|
||
Gankra suggested I update to 0.10.1 which includes code_id fixes. That's b0af5b4c.
|
|
Assignee | |
Comment 3•3 years ago
|
||
|
|
Assignee | |
Comment 4•3 years ago
|
||
This updates rust, updates rust-minidump, switches the repository for it, updates the command line, and also rewrites ModuleURLRewriteRule to work for the new world.
I spot-checked 20 crash reports. The following are changes I saw:
Differences:
- old version emitted partially uppercased code ids like
53D574219c000, new version emits code ids that are all lowercased like53d574219c000--Gankra says this is from code_id fixes for 0.10.1 - order of modules changed
- new version has threads.N.frames.N.unloaded_modules, though all the values are
null - new version has crashing_thread.frames.N.unloaded_modules--again, all null
- old version sometimes has spaces in system_info.os_ver value; new version does not
- some threads.N.frames.N.module values are now uppercase. e.g. KERNELBASE.dll
Crash report 876ce620-09fe-4df4-af95-b981a0220322 used to have code_id like "0000000056000" and now have "ad492742f884386ba450fac281b9ffa4" which is also the debug_id. I don't see in the code where code_id becomes debug_id. Gankra says it's fine, so I guess it's fine.
Crash report 74ff7e6e-8e92-4a36-85f9-545b10220322 switched from debug_id = "" and code_id like = "00000000100000" to debug_id =
"000000000000000000000000000000000" and code_id = "" for all the modules.
Crash report 37aa7507-2d6e-40e5-aa05-1ebe50220322 is similar. It also has some differences in registers:
| field | old value | new value |
|---|---|---|
| crashing_thread.frames.0.registers.r14 | 0xf57d484f | |
| crashing_thread.frames.0.registers.r13 | 0xc617eb90 | |
| crashing_thread.frames.0.registers.r15 | 0xf42217d0 | |
| crashing_thread.frames.0.registers.r11 | 0x13b19c30 |
Crash report e0a8a338-7ff6-4746-ac48-3affe0220322 had lots of modules with code_id "00000000000000000000000000000000" that changed to "". That looks like a good change.
20 crash reports isn't much, but I don't have a good way to look at more without my eyes catching fire and falling out of my head. Gankra has a socc-pair tool, but that doesn't quite do what I need it to. I started a socorro-stackwalker-utils repo (https://github.com/willkg/socorro-stackwalk-utils) that I'm going to collect things in. Maybe the tools will merge down the road. Maybe not.
|
|
Assignee | |
Comment 5•3 years ago
|
||
|
|
Assignee | |
Comment 6•3 years ago
|
||
This deployed to stage, but it's erroring out with this:
error: The argument '--symbols-url <symbols-url>' requires 1 values, but 4 were provided
USAGE:
minidump-stackwalk [FLAGS] [OPTIONS] <minidump> [--] [symbols-path]...
For more information try --help
(Also, hooray for the debug tab and better error reporting!)
I wrote up an issue in rust-minidump: https://github.com/rust-minidump/rust-minidump/issues/518
Right now this is blocking prod deploys. If we need to deploy something, we can revert this change, deploy, and then re-apply this later.
|
|
Assignee | |
Comment 7•3 years ago
|
||
|
|
Assignee | |
Updated•3 years ago
|
|
|
Assignee | |
Comment 8•3 years ago
|
||
|
|
Assignee | |
Comment 9•3 years ago
|
||
This went to production in bug #1763234 just now. This should fix the missing symbols issues and everything else. Yay! Marking as FIXED.
Description
•