Sourcemaps are broken on Webpack 2 when you use ES6 imports

[gaearon]

See here: https://twitter.com/zachcodes/status/835233480640499713.
Would be nice to verify we're not affected before shipping Webpack 2 to our users.

cc @zackify for the report
cc @benvinegar in case he has any context on Sentry side of things

[benvinegar]

Someone opened the same issue on getsentry/sentry-javascript#844 a few weeks ago, but didn't provide any context. I'm going to assume it's a real thing.

[benvinegar]

Oh, my bad, they just suggested Raven.js doesn't work at all, which sounds false. Given they didn't follow-up I'm going to just close it.

[zackify]

This seems to be okay. I was using hidden source maps and the issue on sentry I was referring to was from an old android device, I'm assuming that may have had something to do with it. I will say, this is super confusing:

devtool: 'source-map',
plugins: {
...
new webpack.SourceMapDevToolPlugin(options),
new webpack.optimize.UglifyJsPlugin({
      sourceMapFilename: 'sourcemap.js.map',
      sourceMap: true,
      ...
    }),

Not a fault of this project, but webpack has three different options to generate a sourcemap. I'm currently going with devtool: 'source-map', and

new webpack.optimize.UglifyJsPlugin({
      sourceMapFilename: 'sourcemap.js.map',
      sourceMap: true,
})

Before, I was trying to export a sourcemap, upload it to sentry, and not host it publicly. Got it working by just using that config and keeping them all public. Hopefully this can help somebody.

[Timer]

Just tested this, source maps we're generating definitely are incorrect (wrong line numbers). This is easily seen in ChromeDevTools.

Same results in error overlay

(and no the exception isn't at that line by some magical compilation, see img below)

[Timer]

Was going to file a bug with webpack, but it appears it was already filed: webpack/webpack#3579.

edit: seems to be a bug in https://github.com/webpack/source-list-map, I recall there being a bug in this module before 😁

[gaearon]

sigh

[zackify]

I have actually seen this myself but didn't think much of it. When developing the error seems to be above the actual line number but I can still figure it out. I haven't seen this happening when I build for production though.

…

On Wed, Mar 1, 2017 at 07:59 Dan Abramov ***@***.***> wrote: sigh — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1638 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAbacGLK-e2W8_k6UXCDpgRww0kMK30Tks5rhWuYgaJpZM4MLrWX> .

[Timer]

@gaearon 0.10 plot twist: https://github.com/motion/pundle 🤷‍♀️

edit: or rollup or brunch or insert new bundler here

[Timer]

On a serious note, I wonder how far pundle has come along since the first PR from @steelbrain making a prototypical switch. Maybe worth looking at -- seems like they're missing a file-loader though.

[steelbrain]

@Timer Pundle has had a LOT of progress since that first PR. We now have boot from cache so projects really big boot up in 3-4 seconds instead of several minutes, we have a matured url loader and still a very stable and fast HMR. We're working on docs right now and are planning to release v2 soon

[Timer]

Does pundle support some sort of file loading? e.g. SVGs that use fragment identifiers?

[steelbrain]

At the moment it supports loading them via base64 inlining (through pundle-loader-url, loading them via links is on the todo list)