Increase of EnterJit crashes on 2025-08-05.
Categories
(Core :: JavaScript Engine, defect)
Tracking
()
People
(Reporter: nbp, Unassigned)
References
(Blocks 2 open bugs, Regression)
Details
(Keywords: regression)
Bug 858032 was crashing at a rate of ~500 crashes per day on July 2025, and ~800 on September 2025.
The inflection point appears simultaneously on release and esr at the date of 2025-08-05.
Crash stats info from July 1st to August 1st:
Product
| Rank | Product | Count | Percentage |
|---|---|---|---|
| 1 | Firefox | 19733 | 66.59 % |
| 2 | Fenix | 9269 | 31.28 % |
| 3 | Thunderbird | 359 | 1.21 % |
| 4 | Focus | 273 | 0.92 % |
Platform
| Rank | Platform | Count | Percentage |
|---|---|---|---|
| 1 | Windows NT | 12757 | 43.05 % |
| 2 | Android | 9542 | 32.20 % |
| 3 | Linux | 4421 | 14.92 % |
| 4 | Mac OS X | 2914 | 9.83 % |
Cpu arch
| Rank | Architecture | Count | Percentage |
|---|---|---|---|
| 1 | amd64 | 15745 | 53.13 % |
| 2 | arm64 | 9578 | 32.32 % |
| 3 | x86 | 4157 | 14.03 % |
| 4 | arm | 154 | 0.52 % |
Crash stats info from September 1st to November 1st:
Product
| Rank | Product | Count | Percentage |
|---|---|---|---|
| 1 | Firefox | 37343 | 82.19 % |
| 2 | Fenix | 7354 | 16.19 % |
| 3 | Thunderbird | 561 | 1.23 % |
| 4 | Focus | 179 | 0.39 % |
Platform
| Rank | Platform | Count | Percentage |
|---|---|---|---|
| 1 | Windows NT | 29357 | 64.61 % |
| 2 | Android | 7533 | 16.58 % |
| 3 | Linux | 5627 | 12.38 % |
| 4 | Mac OS X | 2920 | 6.43 % |
Cpu arch
| Rank | Architecture | Count | Percentage |
|---|---|---|---|
| 1 | amd64 | 33537 | 73.81 % |
| 2 | arm64 | 7831 | 17.23 % |
| 3 | x86 | 3899 | 8.58 % |
| 4 | arm | 170 | 0.37 % |
The date does not corresponds to any releases.
The reported URL do not highlight anything significant.
|
||
Comment 1•8 months ago
|
||
ESR and release increasing at the same time makes me suspicious. There were no ESR dot releases near that date; we released three ESRs (115, 128, and 140) on July 22, and then nothing until Aug 19. (eg 140.1.0, 140.2.0). We did have a couple of dot releases for 141 around that time (141.0.2, 141.0.3), but they don't seem to include any SM changes.
Could this be due to a change in how we categorize crash signatures? That would explain why it affects multiple channels at the same time without any code changes.
|
Reporter | |
Comment 2•8 months ago
|
||
Maybe Gabriele would know if there is any infra changes on our side which can explain this sudden change?
|
||
Comment 3•8 months ago
|
||
Yes, we updated the stackwalker tool used by Socorro around that date (see bug 1973186). It's possible that it started generating crashes with different stacks and landed more crash reports here. I could verify it by running the new and old versions of the stackwalker on a bunch of crashes from this signature, but I don't have many spare cycles at the moment.
|
|
Reporter | |
Comment 4•8 months ago
|
||
(In reply to Gabriele Svelto [:gsvelto] from comment #3)
Yes, we updated the stackwalker tool used by Socorro around that date (see bug 1973186). It's possible that it started generating crashes with different stacks and landed more crash reports here. I could verify it by running the new and old versions of the stackwalker on a bunch of crashes from this signature, but I don't have many spare cycles at the moment.
Thanks for checking.
Given the perfect match of the date and the fact that nothing else seems relevant, I guess this would not be necessary.
Especially given the list of changes, which is exactly what we are noticing here, i.e. more Windows x64 entries for JIT frames.
Description
•