2020h1 was rough and 2020h2 was not to be outdone. 2021h1 was worse in a lot of ways, but I got really lucky and a bunch of things happened that made 2021h2 much better. I'll talk a bit more about that towards the end.
But this post isn't about stymying the corrosion of multi-year burnout--it's a dizzying retrospective of Socorro engineering in 2021.
Highlights of 2021
This is long.
Fixed bugs in Breakpad minidump-stackwalk (C++, bash)
I fixed a handful of issues in minidump-stackwalk. I fixed printing addresses to be left-padded so they sort better, improved project scaffolding and scripts, and documented the JSON output.
I also reviewed several pull requests from gsvelto and gankra for adding support for printing
winerror.herror codes, printing
__fastfailerror codes, thread names, interpreting
NTSTATUSerror codes, listing unloaded modules, surfacing the
LastErrorValue, better support for AArch64 when encountering buggy CFI information, and support for AMD64 and X86 variable-sized contexts.
Ended email address collection
Since the beginning of crash collection at Mozilla, we've asked users to add their email address to crash reports. Early on, this was really helpful as it allowed us to contact users and get more information about what was going on when Firefox crashed.
Some crash reporter dialogues had an email field in the form, but many did not. Further, crash reporter dialogues in other products did not have an email field. I wanted to either fix all the forms so they all had an email field or end collection for it.
In 2020, I worked on figuring out whether it was still useful. I talked to people across the company: product managers, engineers, people involved in stability efforts. However, I wasn't able to build a compelling case to keep it.
Email addresses are identifiable information. Without a compelling case to collect and store email addresses, we should stop collecting and storing it.
In 2021, I added the email field to the list of fields we remove from crash reports before ingestion. I removed all email-field-related bits from the processor and Crash Stats websites. I also removed the email field code from Firefox crash reporter dialogs. This was incredibly difficult since that code is ancient, has no tests, and is pretty fickle. For example, the macOS dialog is implemented in an old
.nibformat which xcode doesn't support so there's no way to edit it until we rewrite the crash reporter dialog.
This work was done in [bug 1688883].
Removed derogatory language from repositories
I switched the main branch to "main" in all Socorro-related repositories and removed/replaced derogatory language.
Added support for Breadcrumbs and structured JavaException annotations
The android-components library supports sending crash reports to both Sentry as well as Socorro and this is the library that Mozilla's Android products use. The code for sending crash reports to Sentry sends breadcrumbs data with crash reports. Breadcrumbs can add context to crash reports which can help in debugging.
I worked with Roger to define a structured form for breadcrumbs information, include it as the
Breadcrumbscrash annotation, and surface it in the report view for crash reports on Crash Stats.
Fennec and Fenix both sent crash reports for Java crashes. Instead of including a minidump which Socorro can process to get a symbolicated stack of the crash, crash reports from Java processes send a
JavaStackTraceannotation which has an unstructured string representing the stack of the crash. Because it's unstructured, it's difficult to do anything with. We can't display it with additional information. We can't use it for improving signature generation for Java crashes.
I worked with Roger to define a structured form for Java crashes and we put this in the new
JavaExceptionannotation. It's displayed in the report view. In the future, we will use this for overhauling signature generation for Java crashes.
I want to continue to improve crash ingestion for Java crashes and Android products.
Do you work on Android products? What can I do to improve crash ingestion and Crash Stats for you?
Overhauled Elasticsearch crash storage
Socorro's processor indexed arbitrary crash data not specified in a schema in Elasticsearch in a way that was unusable, took up storage space, and made it difficult to know exactly what data we were storing. Further, this created type errors when adding new things to the index that were difficult to work around.
I reworked the Elasticsearch crash storage code to only index crash data specified in a schema. This reduced storage use, made it easier to know what was and wasn't in the index, and completely eliminated type errors when adding new fields.
This also gets us closer to being able to either upgrade Elasticsearch to a more recent version or migrate off of it.
Improved symbolication and signature generation
The ultimate goal is to support symbolication and signature generation for data sets outside of crash reports in Socorro. This includes address information in the Firefox profiler, symbolicating and generating signatures for crash ping data, symbolicating and generating signatures for crashes in Firefox CI, etc.
I've been working on this for several years now. The systems and libraries involved are all in active use and the project is huge, so progress is in small, incremental steps. I made a lot of progress in 2021 even if it doesn't look like it.
I hacked some fixes in fx-crash-sig to unblock the graphics and Fission teams trying to use the crash ping data. I started on a rewrite to improve the API making it easier to use especially for symbolicating and generating signatures for large batches. I might be able to get to that in 1. I pushed out 3 releases this year.
I fixed a variety of issues in socorro-siggen. It still requires manual syncing with the signature generation code in Socorro. I pushed out 4 releases.
I completed and pushed to production the new Mozilla Symbolication Service. The new service is built on top of Symbolic which Mozilla is increasingly using in our crash reporting ecosystem. This supports line numbers and other file data and set us up for possible future support of inline functions and other debug info files. It also allows us to optimize performance specifically for symbolication without having to worry about the other responsibilities of the Mozilla Symbols Server. It's also a big step in the effort of Rust-ifying crash reporting. I plan to write up a project retrospective for this in 2022.
I also finished an alpha-quality API for signature generation. I want to improve the API to make it easier to use for data sets that aren't crash reports and crash pings. That will likely lead to backwards incompatible changes to the API. I'm planning to finish this up in 2022. You can see what we've got at https://crash-stats.mozilla.org/api/#CrashSignature.
Established Crash Reporting Working Group
Unlike Telemetry where the bulk of the people working on the client, ingestion, tooling, and analysis for Telemetry data are all in Data Org, the wild wild world of crash reporting is all over Mozilla. It feels chaotic to me and I keep getting surprised by initiatives that need crash ingestion support. Further, we've got a lot of historical knowledge, the documentation is all over the place and isn't great, it's hard to find support for crash reporting related problems, and it's probably the case that if we collectively focused on specific initiatives, we could achieve so much more.
With that, I figured I'd create a Crash Reporting Working Group and see if we could use that to pull everything and everyone together.
I was all over the place this years, so I didn't spend as much time on this as I wanted to. With a lot of help from Gabriele, I think we improved the situation in a few key ways:
we renamed #breakpad to #crashreporting on Matrix and pushed that as the singular place for support for crash reporting, Crash Stats, symbols, etc
we did an inventory of projects, teams, and owners for crash reporting related things making it easier to find people
I started monthly newsletters detailing changes in crash reporting ecosystem
Gabriele ran a Crash Reporting All Hands giving stakeholders a chance to see what's going on and bring up their issues which we could factor into ongoing development efforts
In 2022, Gabriele wants to focus on updating, consolidating, and improving the documentation around crash reporting. Towards this, I want to work on the crash ingestion side of that and it'd be great to finally build out data documentation for crash annotations similar to what Telemetry has with the Glean Dictionary.
Unified conventions and standards across the Socorro-verse
The crash ingestion pipeline and symbols services are composed of a myriad of repositories, libraries, infrastructure code, tooling, and other things. It's hard on me when things are different across them all. Some of them I built, but many of these pieces I inherited from other individuals and teams over the years.
In 2021, I worked on unifying the pieces that affect development and maintenance workflow. Things like Dockerflow setups, scripts, makefiles, testing harnesses, documentation tooling, release tooling, bug systems, metrics, and system test tooling.
I'm reworking all of the service components to use Everett for configuration. This makes it a lot easier to write configuration-using code, tests, and documentation.
In doing that, I reduced a lot of friction I have when switching between pieces of the project. That's helped. There's still some other things I want to do, but I'll get there when I get there.
I'm keeping an eye on the various initiatives to standardize technology and architecture for services at Mozilla. As those initiatives move towards more concrete conventions, standards, and guidelines, I'll update Socorro-verse things accordingly.
Fixed a bug in dump_syms (Rust)
I fixed a bug in dump_syms to normalize "anonymous namespace" symbols which fixed an issue with signature generation in Socorro. It wasn't a big bug, but it was pretty straight-forward to fix. I like working in Rust.
This was done in pr 192.
Updated socorro-submitter to improve parity with collector
Over the last couple of years, we've made changes to how the crash reporting client submits crash reports to the Socorro collector. Socorro has a socorro-submitter that resubmits crash report data from the prod to stage environment.
I finally got around to fixing the collector and socorro-submitter to submit crash reports from prod to stage in the same way they were submitted to prod. That helps us suss out issues in the crash ingestion pipeline in our stage environment.
Crash data improvements
I rewrote the macOS version normalization code to correctly follow the Mac platform OS names across the different versions.
I rewrote the Windows version normalization code in the same way handling Windows 11 builds.
I added a
major_versionfield which is indexed making it easier to search for all crash reports for a specific major version. This was a feature request that came up pretty often. [bug 1111612]
We added support for
mac_crash_infoextracted from the minidump, though it's not particularly great.
We added support for
I added a
crash_report_keysfield which is a list of all the crash annotation names and names of dumps that came in the crash report. This allows us to search for crash reports that included a specific file or annotation. This makes adding support for new annotations a lot easier because it'll be easier to find example crash reports. Gabriele says this also allows us to deprecate unused annotations, too--something that's been very difficult to do until now.
Improved report view in Crash Stats
Personally, I hate the report view in Crash Stats. I find it disorganized, confusing, and ugly. I have extremely strong feelings about this for whatever reasons. I blame myself.
Anyhow, in 2021, I made some minor improvements:
Redid what is now the Crash Annotations tab so it includes the original values of the crash annotations that came in the crash report. It's also clearer which fields are public and which are protected data. I also tried to break out the metadata that the crash ingestion pipeline adds to the crash report so that's clearer, but that's still a work in progress.
Redid what is now the Raw Data and Minidumps tab so it has more convenient links to artifacts we're storing (raw crash, processed crash, dumps, memory report, etc) and is clearer about what's protected data. I still don't like this tab, but it's "better".
Improved the Extensions tab in Crash Stats so it shows the signed state.
Fixed the Correlations tab so it always shows up with either correlations data or a helpful message even if correlations aren't available.
I also implemented a Debug tab which I wish I had thought of years ago. The Debug tab is the magical place where we can collect all the information about how a crash report was ingested so as to debug ingestion and stackwalking issues. It's amazing and has already been incredibly useful. In 2022, I'll add signature generation notes.
Wound down the tools-socorro mailing list
Before I started working on Socorro, it was an open source project and had multiple companies running their own Socorro installation. As I understand it, this was coordinated using the tools-socorro mailing list.
When I joined Socorro, that list was pretty dead and we had no contact with anyone running their own Socorro.
I refocused the Socorro project so it was the code that ran the Mozilla crash ingestion pipeline and ended supporting other people running their own crash ingestion pipelines. I've talked about this before.
Anyhow, the last step here was to wind down the tools-socorro mailing list.
I improved argument handling for crashstats-tools and redid
supersearchfacetso it supported other periods.
So now you can do something like this:
$ supersearchfacet --_facets=product --period=daily date Fenix Firefox FirefoxReality Focus ReferenceBrowser 2021-12-14 00:00:00 55723 101260 125 1598 0 2021-12-15 00:00:00 50878 111818 202 1492 1 2021-12-16 00:00:00 51651 101585 129 1308 0 2021-12-17 00:00:00 51688 99147 121 1353 0 2021-12-18 00:00:00 54984 59396 97 1490 0 2021-12-19 00:00:00 54076 55369 202 1582 0 2021-12-20 00:00:00 52187 81295 127 1560 0 2021-12-21 00:00:00 50243 74869 99 1381 0
crashstats-tools are written in Python, but I'd like to rewrite them in Rust. Users of crashstats-tools are predominantly running Rust command line tools and it'd be a lot easier for them to install and use if written in Rust.
Consulted on crash reporting
Gabriele is the go-to person for helping with crash reporting client work and I'm the go-to person for the crash ingestion and analysis side.
I consulted on:
Analysis on Fission crashes
This was tricky because Fission crashes aren't like main and content process crashes, so we can't pop up a crash reporting dialog. All the options kind of sucked. I think they eventually went with looking at Crash Stats for what it had and ad hoc analysis of crash ping data for the bigger picture view.
I wish we could have done better here, but I don't have the required bits set up, yet.
Analysis of top crashers
Crash report data isn't representative of crashes our users are experiencing. The graphics team was working hard on webrender and needed a better data set representative of what our users were experiencing. I fixed issues with fx-crash-sig and socorro-siggen allowing them to put together an extremely impressive top crashers dashboard.
MozillaVPN is implementing crash reporting. I worked with Marcus to set up Crash Stats and debug issues with crash submission. (As of this writing, I think he's got it working). While we were debugging, I improved the documentation for crash reporting clients to be more helpful for projects that aren't using Breakpad tooling.
Consulting helped improve documentation and support for other crash reporting clients.
If you're a product manager and starting a new product, please get in touch with me early on to make your crash reporting story a lot easier for all of us!
I redid the architecture and flow diagrams for Socorro and Tecken. This is the first major update to the diagrams we've done in several years. This was an important thing to do before figuring out the AWS to GCP migration.
Previously, I was using graphviz dot notation because it's a text-based source that I can manage and maintain with text-based tools like git, grep, and text editors. However, it's really finicky to render and I spent a lot of time tweaking the diagrams so they rendered in a readable way.
Jeff said we're switching to mermaid.js and diagrams.net. I had the same problems with mermaid.js that I did with graphviz/dot. I tried out diagrams.net and if I save the diagram as a SVG and export as PNG then I can do all the things I want to do and I get a muuuuuuuch better diagram out of it.
I documented the JSON schema for minidump-stackwalk output. This turned out to be really helpful when switching to the new Rust minidump stackwalker. This also helped us remove some code in the Socorro processor for fields that were redundant or no longer used.
I rewrote the Overview documents for Socorro and Tecken after reading ARCHITECTURE.md blog post.
I switched to capturing architecture decisions with ADRs and then backfilled ADRs for Socorro and ADRs for Tecken. This is in line with what other parts of Mozilla are now doing. Plus it'll help me remember what past me was thinking. I have some more ADRs to backfill. I'll continue working on this in 2022.
I improved the Specification: Submitting crash reports document to focus it on people writing crash reporting clients that aren't based on the Breakpad upload tool or our existing crash reporting client machinery.
I did a Socorro Overview: 2021 presentation. In order to do that, I had to write a thing to export my slides into a blog post form.
I updated infrastructure documentation and runbooks for all the services I maintain.
I overhauled the documentation in Mana for Socorro and Tecken and pulled it into the Data Org space.
Crash ingestion involves handling of category 4 data (see Data Collection). One of my jobs is thinking about what we're collecting, how it's accessed, and what happens to it.
With help from Nneka and Emily, I honed our data collection, access, and retention policies. I started a Data Sharing Decision log as well. This will help us be more consistent even as staff come and go.
It's great that this exists, but one of the problems the Socorro-verse has is that information is in a bunch of different places and that's confusing and makes things difficult to find for anyone who isn't me.
In 2022, I'll be working on straightening out the crash ingestion documentation story. I'm keeping tabs on other standards and conventions at Mozilla that I can adopt to make things more consistent.
Ended collection of FennecAndroid, Thunderbird, and SeaMonkey crash reports
FennecAndroid is the version of Firefox for Android before the Fenix rewrite. We have migrated users to the new version and we no longer support the old version. I contacted product managers to establish that this data no longer helps us and then ended collection.
I've been working on ending collection for Thunderbird and SeaMonkey for several years. This was tougher because these two projects are still in use and I needed to help both projects migrate to other services. That took a while.
Switched to new Rust-based minidump-stackwalker
Turns out this was the year!
Aria, with help from Gabriele, Ted, and others, worked on rust-minidump and achieved parity with the Breakpad minidump-stackwalker the Socorro processor was using.
I implemented build tooling, a processor rule to run the new stackwalker, a feature flag to let us switch between the two stackwalkers, several test harnesses to validate the new stackwalker, and redid some parts of the processor to make it easier to observe and debug the new stackwalker in a server environment. This was the impetus behind the new Debug tab.
After a flurry of work, we did the final migration switching the production environment to the new stackwalker last week. CPU and memory usage metrics on the processors went down. Several errors we were getting vanished. It was a perfect deploy.
Check out the average timing for the stackwalker rule drop by half:
The stackwalker rules take up the vast majority of the time it takes to process a crash report so switching to the new Rust-based stackwalker effectively halves the amount of time it takes to process crash reports.
I plan to write up a project retrospective on this in 2022.
I also did a bunch of small features, signature generation changes, reprocessing, bug fixes, docs fixes, and other things.
Other non-Socorro-verse things:
Reviewed Bleach pull requests and re-upped as co-maintainer
I stepped down from Bleach in March 2019, but I continued to help out with reviewing changes. Bleach development has slowed way down, but we had one security issue and a handful of other things to work through.
In December 2021, I stepped back up as co-maintainer. I started working through issues and pull requests. I think I want to keep Bleach maintenance to a minimum and only work on things that are urgent and/or necessary.
Switched to Jupyter for analysis
I switched to using Jupyter for data analysis. For a long time, I've been looking for a way to capture data and analysis such that it stuck around and was viewable long after the fact. Socorro crash report data expires after 6 months, Grafana data expires after 6 months, logs expire, etc.
GitHub can show notebooks, so I discovered if I save the notebook and data files, I can do my analysis in Jupyter notebooks and save it in a way that doesn't expire and allows other people to tinker with the data.
Updates to tooling
I made some minor fixes and releases for rob-bugson.
I overhauled paul-mclendahand to require fewer weird configuration settings and setup and support GitHub access tokens.
I improved my socorro-release script which I use for creating deploy bugs and tagging releases for services I maintain.
I overhauled Everett configuration components and got about half-way through redoing the Sphinx extension for auto-generating configuration documentation. (The new Sphinx extension changes are amazing. I really wish I had time to finish it up.)
I did some minor updates to Markus. Generally, Markus is "done", so I'm only putting out updates to improve documentation and support new Python versions.
Joined Data Stewards
I joined Data Stewards and did a handful of data reviews.
One of my reasons for joining was to unify Socorro's crash annotation processes with Telemetry's data processes. I haven't really gotten very far with that, yet. Maybe next year.
Started the Data Neighborhood
I had a regular 1:1 with Will and during one of these conversations, we were talking about how I went on PTO for 5 weeks and he took a 5 week sabbatical at the Recurse center and when we came back, it was so quiet. We thought about what we could do or facilitate to increase the activity, interconnectedness between people, and general feeling of belonging. It's hard to describe the project well in a short paragraph, so I'll leave that for another blog post.
Anyhow, we started the Data Neighborhood as a way to provide a framework for organizing and running activities, events, groups, etc.
Out of that, we started a
#data-checkinsSlack channel where people could talk about their day. We started a
#data-writing-helpchannel for talking about writing which is a critical part of our work and also getting help with editing and review. We put together a Mana page that listed many groups across Mozilla and where they congregate (they're hard to find--there's no index anywhere).
I was talking with Mark about this and my current thinking is that it was a good experiment, but it didn't ultimately help much because the hypothesis of "if we provide a framework for people to do things, they'll do them!" didn't pan out.
Also, now that Will is leaving, I'm not sure what I'm going to do with this. Should I pull in some other people and see where that goes? Should I shelve it? What's the next hypothesis we can test out?
In September, I hit my 10-year Moziversary.
5 weeks of PTO
I work on Socorro-verse and a bunch of other stuff on my own. It's against my nature to step away for periods of time in situations when I'm on my own. Who's going to press the button every 108 minutes? What if the island blows up while I'm not there?
But I was incredibly burned out and had been for years and years for all the reasons.
I talked it out with Mark, my manager, who helped me pull off a 5-week PTO. During that time, I did nothing. I made no plans. I didn't work on projects. I didn't take any trips. I didn't take classes. I didn't read books. I did nothing.
Coming back was so totally weird. I felt like a different person. I look over this description and it barely evokes how stark the before and after was.
I'm still feeling weary and burned out, but I feel much better.
I've never taken that much PTO off at once in my life. Previously, I would quit, take the time off as needed (one time I took a year off), and then join a different company. I feel really lucky to have been able to take 5 weeks off and come back to an environment where my colleagues were actively looking out for my well-being.
"Attended" some conferences
I attended PyCon 2021, but I also had my second vaccine shot, so I slept through the last day.
I attended the StaffPlus Live conference which was fascinating. As a new Senior Staff Engineer, I found this conference really helped me frame my work and role in the things going on around me for my new role plus it helped frame the last few years I spent as a Staff Engineer. I had no idea about all the stuff I had no idea about. Also, Will Larson's Staff Engineer: Leadership beyond the management track really helped me up my game.
I attended Google Next 21. I know they put a lot of work and effort into it, but it's such a dizzying alphabet soup of service stuff it's hard to follow.
Stuff I worked on, but either abandoned or put on the back burner:
Folding in socorro-siggen into socorro again
Overhauling crash signature generation for Java crashes
Symbolication and signature generation for crash ping data
fx-crash-sig overhaul for better batch support
Project planning for a possible Socorro-as-a-service
Plan for unifying Socorro with Telemetry ingestion pipeline
Overhauling crash report data structure and building schemas for crash reports and processed data
Supporting debug files other than Breakpad sym
Supporting inline functions in symbolication
Retrospective for 2021
This post is ridiculously long. When I started writing it, I was of the mind this year just kind of flew by without much progress and was another let-down of a year. Now that I've written most of the blog post, I think that's probably not the case.
I wish I had gotten further on some projects--especially sorting out crash pings and making that data set more usable. I also wish I hadn't started so many different things. One of the things I tend to do is move a lot of projects forward in short, incremental steps towards various goals while keeping everything working. There's a lot of overall progress, but not a lot of individual progress and some projects don't hit the jackpot until the very end when they're finished.
But whatever--I got lots of feelings. I think I done good in 2021.
Many thanks to Mark who looks out for me and makes it easier for me to be me and me to do the things I do. I wish everyone had a Mark or a Mark-like. I would eagerly pay for Mark-as-a-Service!
Many thanks to Jason who's a first-rate SRE!
Many thanks to Aria, Gabriele, Jeff, Markus, Calixte, Roger, and Ted whose efforts on Rust-ifying everything and knowledge of crash reporting and engineering minutia are critical to any success I might ever have with anything!
Many thanks to Chris, Jim, Nika, and others who rely on Crash Stats and crash ingestion for their patience as I muddle along and their insight as to how they need things to work to do the things they need to do!
Many thanks to Lonnen whose historical knowledge helps give context to the eldritch weirdness!
Many thanks to Will, Jeff, John, Rob, Paul, Chutten, Les and everyone else who make up my peer support group!
Many thanks to you, dear reader, for getting through this omnibus! I hope it was helpful to you.
Here's some Bugzilla and GitHub numbers because it's easy to build Bugzilla and GitHub numbers and sometimes they're interesting.
Bugs created: 311 Creators: 38 Will Kahn-Greene [:willkg] ET : 219 Gabriele Svelto [:gsvelto] : 30 Sebastian Hengst [:aryx] (need : 7 Steven Michaud [:smichaud] (Re : 6 Andrew McCreight [:mccr8] : 5 Markus Stange [:mstange] : 4 Aria Beingessner [:Gankra] : 3 [PTO until Jan 10th] Agi Sferr : 2 Jeff Muizelaar [:jrmuizel] : 2 and others ... Bugs resolved: 316 WONTFIX : 22 FIXED : 268 MOVED : 2 WORKSFORME : 9 DUPLICATE : 5 INCOMPLETE : 5 INVALID : 4 Resolvers: 14 Will Kahn-Greene [:willkg] ET : 295 Gabriele Svelto [:gsvelto] : 10 Calixte Denizet (:calixte) (in : 2 continuation : 1 mcastelluccio : 1 and others ... Commenters: 107 willkg : 1472 gsvelto : 93 smichaud : 44 peterbe : 35 continuation : 21 and others ... Statistics Youngest bug : 0.0d: 1685461: Upload symbols for the 84.0.1 and 84.0.2 builds... Average bug age : 265.1d Median bug age : 8.0d Oldest bug : 2626.0d: 973894: Supersearch: field exists should be available f...
mozilla-services/socorro: Merged PRs: 221 Committers: willkg : 210 (+13324, -9110, 285 files) dependabot[bot] : 4 ( +52, -48, 5 files) jrmuizel : 2 ( +2, -0, 2 files) jcristau : 2 ( +2, -1, 2 files) gabrielesvelto : 1 ( +26, -1, 2 files) kbrosnan : 1 ( +5, -2, 1 files) amccreight : 1 ( +2, -0, 1 files) Total : (+13413, -9162, 285 files) Most changed files: docker/Dockerfile (36) requirements.txt (32) requirements.in (30) socorro/processor/rules/mozilla.py (28) socorro/external/es/super_search_fields.py (28) webapp-django/crashstats/crashstats/jinja2/crashstats/report_index.html (22) socorro/unittest/processor/rules/test_mozilla.py (21) webapp-django/crashstats/crashstats/models.py (20) socorro/processor/rules/breakpad.py (18) webapp-django/crashstats/crashstats/tests/test_views.py (16) Age stats: Youngest PR : 0.0d: 5948: bug 1746872: add note about upload_file_minidum... Average PR age : 0.4d Median PR age : 0.0d Oldest PR : 32.0d: 5802: Bug 1715634 - add get_fpsr to the irrelevant si... mozilla-services/antenna: Merged PRs: 41 Committers: willkg : 37 ( +3114, -2744, 55 files) dependabot[bot] : 4 ( +219, -163, 2 files) Total : ( +3333, -2907, 55 files) Most changed files: requirements.in (20) requirements.txt (20) docker/Dockerfile (13) antenna/throttler.py (7) Makefile (5) antenna/app.py (5) antenna/breakpad_resource.py (4) systemtest/conftest.py (2) systemtest/test_dockerflow.py (2) antenna/ext/s3/connection.py (2) Age stats: Youngest PR : 0.0d: 763: fix reqs rules to not load dependencies Average PR age : 0.0d Median PR age : 0.0d Oldest PR : 0.0d: 763: fix reqs rules to not load dependencies mozilla-services/tecken: Merged PRs: 87 Committers: willkg : 81 (+12635, -8986, 153 files) dependabot[bot] : 6 ( +19, -19, 4 files) Total : (+12654, -9005, 153 files) Most changed files: requirements.in (30) requirements.txt (30) docker/Dockerfile (18) eliot-service/eliot/symbolicate_resource.py (11) frontend/yarn.lock (11) eliot-service/tests/test_symbolicate_resource.py (9) eliot-service/eliot/app.py (7) eliot-service/eliot/cache_manager.py (7) docs/overview.rst (5) Makefile (5) Age stats: Youngest PR : 0.0d: 2460: Update symbolicate script to use new service Average PR age : 0.1d Median PR age : 0.0d Oldest PR : 3.0d: 2397: bug 1673887: rewrite disk cache manager mozilla-services/socorro-submitter: Merged PRs: 6 Committers: willkg : 6 ( +908, -581, 25 files) Total : ( +908, -581, 25 files) Most changed files: requirements-dev.txt (4) requirements.txt (3) src/submitter.py (3) docker/test/Dockerfile (2) tests/test_submitter.py (1) .gitignore (1) Makefile (1) README.rst (1) bin/aws_s3.sh (1) bin/build_artifact.sh (1) Age stats: Youngest PR : 0.0d: 36: Update urllib3 Average PR age : 0.0d Median PR age : 0.0d Oldest PR : 0.0d: 36: Update urllib3 mozilla-services/minidump-stackwalk: Closed issues: 11 willkg : 2 gabrielesvelto : 1 Merged PRs: 21 Committers: willkg : 10 ( +642, -354, 14 files) gabrielesvelto : 8 (+27998, -42, 176 files) Gankra : 2 ( +188, -3, 2 files) steven-michaud : 1 ( +596, -0, 2 files) Total : (+29424, -399, 185 files) Most changed files: minidump-stackwalk/stackwalker.cc (10) README.rst (4) bin/build_breakpad.sh (3) breakpad-patches/20-winerror-codes.patch (2) .gitignore (2) bin/run_mdsw.sh (2) breakpad-patches/16-get-last-error.patch (2) bin/build_stackwalker.sh (2) bin/clean_artifacts.sh (2) CODE_OF_CONDUCT.md (1) Age stats: Youngest PR : 0.0d: 47: Readme fixes Average PR age : 2.3d Median PR age : 0.0d Oldest PR : 17.0d: 11: Update breakpad and apply the stackwalker impro... willkg/socorro-siggen: Closed issues: 5 willkg : 1 Merged PRs: 13 Committers: willkg : 13 ( +946, -467, 24 files) Total : ( +946, -467, 24 files) Most changed files: setup.py (6) siggen/cmd_signature.py (6) HISTORY.rst (5) siggen/__init__.py (4) .github/workflows/main.yml (4) tox.ini (4) siggen/rules.py (4) siggen/socorro_sha.txt (4) siggen/tests/test_rules.py (4) siggen/siglists/prefix_signature_re.txt (3) Age stats: Youngest PR : 0.0d: 84: Prep for 1.0.8 release Average PR age : 0.0d Median PR age : 0.0d Oldest PR : 0.0d: 84: Prep for 1.0.8 release willkg/crashstats-tools: Closed issues: 4 Merged PRs: 9 Committers: willkg : 9 ( +197, -114, 18 files) Total : ( +197, -114, 18 files) Most changed files: setup.py (3) .github/workflows/main.yml (3) crashstats_tools/utils.py (3) src/crashstats_tools/__init__.py (2) Makefile (2) tox.ini (2) crashstats_tools/cmd_reprocess.py (2) HISTORY.rst (1) README.rst (1) src/crashstats_tools/cmd_fetch_data.py (1) Age stats: Youngest PR : 0.0d: 43: Prep for 1.2.0 release Average PR age : 0.0d Median PR age : 0.0d Oldest PR : 0.0d: 43: Prep for 1.2.0 release All repositories: Total closed issues: 20 Total merged PRs: 398