TargetClosedException after Page.CloseAsync()

First of all, thank you for your tremendous effort that went into this cool project!

I’m trying to generate PDF file, as you will see in the snippet below and although everything works fine, even when I fire multiple concurrent requests that are opening multiple instances of chromium, I’m seeing some exceptions in the log files.

I’ve been able to narrow down that the issue occurs when trying to close the connection. The following exception is being thrown:

• PuppeteerSharp.TargetClosedException: 'Navigation failed because browser has disconnected! (Target.detachedFromTarget)'

Example reproducing the issue:

            Stream stream;
            await new BrowserFetcher().DownloadAsync(BrowserFetcher.DefaultRevision);

            var browserOptions = new LaunchOptions
            {
                Headless = true,
                Args = new[] { "--no-sandbox" },
            };

            var browser = await Puppeteer.LaunchAsync(browserOptions, loggerFactory);
            var page = await puppeteerService.GetPage(browser);

            await page.SetContentAsync("<div>Testing</div>", puppeteerService.GetNavigationOptions());
            stream = await page.PdfStreamAsync(puppeteerService.GetPdfOptions());

            await page.CloseAsync(); // after this line the exception is thrown
            await browser.CloseAsync();

Versions

PuppeteerSharp: 1.12.1 .NET: netcoreapp2.1 Platform: Windows 10

Attached log file (LogLevel.Trace): puppeteer-sharp.log-2019-03-14.txt

[R2D221]

I created a pull request (https://github.com/kblok/puppeteer-sharp/pull/1218)

I think I found the issue. The problem is when disposing an object, if it has unhandled tasks, it goes to the event. So, what I did was take them and handle them myself in the finalizer.

Let me know what you think.

[kblok]

@Emersonbg I was about to suggest that. One thing is the browser opening a big page and another is the browser receiving a massive CDP message. On the other hand, SetContentAsync is no more than an evaluate setting the HTML content of the document. You could implement that and send the content in chunks.