Description
Steps to reproduce:
-
Run my company's React project for debugging:
npm run dev
The service will run, and you can access it at http://localhost:80. -
Modify the
hosts
file to map127.0.0.1
toxxx.yyy.com
.
You can then access the above service by openingxxx.yyy.com
. -
In DevTools, locate the code and add any breakpoint that might be triggered. Then, trigger this breakpoint through an operation.
You will see that the breakpoint is effective. -
For the first 1-2 seconds, DevTools is still responsive, but after that, it becomes unresponsive. You can't switch tabs or inspect variables in the call stack. Additionally, you can see that its CPU usage spikes, consuming an entire CPU core. If you resize the DevTools window, the newly exposed part of the window isn't rendered.
-
If you choose to close DevTools, the program resumes normal operation.
About the Edge version:
This issue did not exist in the initial version 122 of Edge. The problem appeared after I upgraded WebView2 from version 122 to 127.
I consulted with some friends at Microsoft, and they asked if the same issue occurred in Edge. At that time, my Edge version was 122, and I couldn't reproduce the issue. So, I upgraded Edge to version 127, and the problem appeared, just as it did with WebView2 version 127.
About the reproduction rate:
In my code environment, the issue has a high reproduction rate, around 90%.
I tried to reproduce the issue in a simple React project, but I couldn't.
It may require a more complex React project, and it should be running in a React debug service environment.
Unfortunately, I cannot provide my company's code, as it's not allowed to be shared.
Screen from edge devtools:
AB#53420578
Activity
captainbrosset commentedon Aug 23, 2024
Thank you for filing. I will make sure this shows up on our dev team's backlog.
robpaveza commentedon Aug 23, 2024
Hi @gzx-miller , I'm Rob Paveza, an Engineering Manager on the DevTools team.
Without knowing a lot about your project, I wanted to ask about one case in which we've often seen this. Is your build system configured to use something like
eval-source-map
orcheap-eval-source-map
as the source map configuration?Some background: some build systems like Webpack provide this as a mechanism to support Hot Module Replacement. In short, suppose you have a single TypeScript file like this:
A conventional way to bundle and minify this might look like this:
Setting aside the issue of minification, what ends up happening to support HMR with
eval
is something like this:That is approximately semantically equivalent modulo ES modules behavior, and it allows the module to be swapped at runtime by changing the object representing the module.
But this produces a terrible debugging experience. So the folks who did that added source maps to it:
So now you see that the source map is base-64 encoded JSON text encoded inline twice: once within the module itself, and another time within the eval. (I have quite literally seen this.) The source maps are actually germane, as well, because they represent two different scripts: the script containing the
eval
, in which the interior script is just a string; and the eval'd script itself. But the net result is that for each bundle (and each time the modules are hot-reloaded), you have 4 copies of the source code being handled within the script engine.I'm not saying that this is causing your issue -- however, I have seen from an internal team that this can cause quite a bit of CPU thrashing and performance degradation. (I also saw OOM crashes from this). It is a severe problem with the HMR mechanism for large projects, so if possible, I'd like to eliminate this as a cause.
gzx-miller commentedon Aug 26, 2024
Hi, @robpaveza ,
Thank you for your response!
It might indeed be the cause, as I have found code related to
evalSourceMap
in the project:However, what methods can I use to bypass or fix this issue?
I have tried remove those code in webpackDevServer.config.js, but the problem still exists.
Currently, this problem is somewhat hindering common debugging work.
Many issues that could usually be inspected by setting breakpoints now require investigation through added logging.
The last question is "Why chrom and edge122 without problem?"
gzx-miller commentedon Oct 21, 2024
I recently found that a necessary operation to trigger this phenomenon is, after hitting a breakpoint, hovering the mouse over the variable to observe its value, which will cause a hang and high CPU usage. However, as long as the variable is not inspected, I can step through the code multiple times fluently. @robpaveza @captainbrosset
wertzui commentedon Nov 8, 2024
I have the same issue since some time.
I'm currently on
Version 132.0.2926.0 (Offizielles Build) canary (64-Bit)
HesamKashefi commentedon Nov 13, 2024
My browser hangs on breakpoint too.
This issue has been around for a couple of months.
I have tested on 3 different computers (one of them is a high-end PC).
My code is an angular application.
On the other browsers there is no problem. (Chrome, Firefox).
leahmsft commentedon Nov 13, 2024
Thank you all for the details about this issue. I'll bring this back to the engineering team to help investigate.
nvdweem commentedon Nov 14, 2024
Same issue here, also with an Angular application, also noticed it for several months.
Running in incognito without any plugins enabled doesn't help.
I did notice that disabling javascript sourcemaps also removes the hang (mostly, it then needs to highlight a huge file but that's way faster still), but disabling sourcemaps alltogether isn't really an option.
leahmsft commentedon Nov 14, 2024
@gzx-miller @wertzui @HesamKashefi @nvdweem
Hi all,
Can you please provide the following details to assist the engineering team's investigation:
wertzui commentedon Nov 14, 2024
Windows
Edition Windows 11 Pro Insider Preview
Version Dev
Build 27744.1000
Funktionspaket Windows Feature Experience Pack 1000.26100.32.0
Edge
Version 132.0.2941.0 (Offizielles Build) canary (64-Bit)
How to reproduce:
HesamKashefi commentedon Nov 15, 2024
I think I fixed it!

I just removed all the extensions and tested again.
Everything looks fine now!
Then I installed Angular DevTools and Redux DevTools again. It became slow again but hangs just a about five seconds not like before which was about 1 minute!
Here's a list of my extensions before I removed them.
My Edge version is : Version 130.0.2849.80 (Official build) (64-bit)
My Windows version: Windows 11 23h2 22631.4460
In the past even when I disabled all the extension nothing was different but now that I removed all and just install Angular and Redux DevTools when I disable them, everything is superfast.
It was really annoying to have such hang duration (1 Minute) on an Intel Core I7 14700K - 32G RAM DDR5 - SAMSUNG 990Pro M2 - ....
nvdweem commentedon Nov 15, 2024
Steps to reproduce
ng new SlowApp
scss
and disablessr
ng s
ngOnInit() { debugger; }
in the classDev Tools
(F12)http://localhost:4200
(this will hit the breakpoint immediately)leahmsft commentedon Nov 20, 2024
@HesamKashefi glad to see that you found a fix and it's working again!
@wertzui are you also seeing this with Angular or just in general?
@nvdweem can you try the same fix @HesamKashefi mentioned above and report back if it worked for you too?
nvdweem commentedon Nov 20, 2024
I had seen that workaround, that's why I included
And
In incognito, no plugins are enabled, so sadly the workaround doesn't work for me.
leahmsft commentedon Nov 20, 2024
@nvdweem apologies for missing that. A new version of Edge Stable recently shipped to v131. I noticed that you're seeing this bug on v130. Have you tried updating to the most recent version?
nvdweem commentedon Nov 21, 2024
I think my Edge updated right after posting the version (Version 131.0.2903.51 (Official build) (64-bit)) and I think I've seen the issue pop up afterwards, but I'm not entirely sure. It doesn't always trigger, but just tried a few breakpoints in and outside of InPrivate and those seemed to be interactive quite quickly.
I'll try to pay attention to whether it triggers again in the coming days.
Update: immediately after this message I hit a breakpoint that took ~15 seconds to become responsive again, refreshing the page and triggering it again 'only' takes ~3 seconds.
HesamKashefi commentedon Nov 22, 2024
@nvdweem I removed all of the extensions, they were already disabled when I had the issue. maybe you could try this too.
I'm guessing that maybe it's from installing extensions 'from other sources'
nvdweem commentedon Nov 22, 2024
Also tried it with a new profile, not a single extension installed (other than a disabled Offline Google Docs extension that was already there). The first breakpoint I hit took about 10s to become responsive, the next one was instantly responsive. A breakpoint a few moments later took ~5s to become responsive. Triggering the same breakpoint in my regular profile also takes about 5s.
It's way faster for these two occurrences as compared to my actual profile in the 130 version, 5s is still a long time to wait on nothing, but more of a nuisance than flow-breaking like the 30s+ delays that I've seen before.
Having following breakpoints trigger quickly feels like a cache miss on the slow triggers. I'll have to pay some more attention, but it could be that breakpoints after code-changes are slower because something (map files?) need to be reparsed or something.
If disabled plugins do anything then that sounds like a way bigger problem than just unresponsive breakpoints.
Edit: Tried the same breakpoint in the blank profile as a slow breakpoint in my regular profile when it was slow and it seems that the blank profile stays pretty quick.
JWess commentedon Feb 4, 2025
I had all the same issues described by @nvdweem when debugging my Angular app. I have the AdGuard AdBlocker extension installed. After reading this thread, I turned off protection for localhost within AdGuard, and that completely resolved the delay I was seeing!
peer-kkang commentedon Apr 7, 2025
+1, Running into same issue when debugging Angular application.
Repro scenario:
Angular version: 19