fix(nextjs): Prevent wrong culprit from showing up for clientside error events

[lforst]

We're seeing a weird issue where error events have the wrong culprit containing incomplete-app-router-transaction.

My suspicion os that a transaction is started, then an error happens in between of the transaction being started and the transaction name being updated, and it thus receives the wrong culprit.

We fix this by not starting the transaction with a wrong name in the first place. I dunno why I didn't do this from the beginning.

I didn't fully verify this but it would be the only explanation I can come up with and I think it is nicer this way in any case.

Note that I even saw this happening with router.push transactions where we don't go into the branch where we don't update the name so it's not related to that code path.

[chargome]

Ok we need to stop working on the same issues in parallel 😂

What I found so far that this is mainly an issue on safari mobile (could e.g. be a failing preflight request) and we only filter on this transaction name for type "transaction"

So I guess we can also remove the event processor with this pr!

ref #15411

[lforst]

Oh whoops, I zoned out in the meeting and thought you were talking about not-found errors or something because I heard event processor and forgot that we had the one for transactions. My bad!

I wouldn't remove the transaction event processor because it serves a different purpose than this fix. This fix is about making sure errors receive the right culprit, the processor is about dropping transactions for dropping transactions that never receive the popevent call - for example when the router.pop is called but the actual navigation never completes and the popevent isn't fired.

[chargome]

Ah yeah I missed that this can still be the case then. Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

[lforst]

Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

Considering this is an invariant that should never happen in the first place I'd rather have it show up and cause people to ask questions rather than hide it behind "wrong" data.

[lforst]

We're seeing a weird issue where error events have the wrong culprit containing incomplete-app-router-transaction.

My suspicion os that a transaction is started, then an error happens in between of the transaction being started and the transaction name being updated, and it thus receives the wrong culprit.

We fix this by not starting the transaction with a wrong name in the first place. I dunno why I didn't do this from the beginning.

I didn't fully verify this but it would be the only explanation I can come up with and I think it is nicer this way in any case.

Note that I even saw this happening with router.push transactions where we don't go into the branch where we don't update the name so it's not related to that code path.

[chargome]

Ok we need to stop working on the same issues in parallel 😂

What I found so far that this is mainly an issue on safari mobile (could e.g. be a failing preflight request) and we only filter on this transaction name for type "transaction"

So I guess we can also remove the event processor with this pr!

ref #15411

[lforst]

Oh whoops, I zoned out in the meeting and thought you were talking about not-found errors or something because I heard event processor and forgot that we had the one for transactions. My bad!

I wouldn't remove the transaction event processor because it serves a different purpose than this fix. This fix is about making sure errors receive the right culprit, the processor is about dropping transactions for dropping transactions that never receive the popevent call - for example when the router.pop is called but the actual navigation never completes and the popevent isn't fired.

[chargome]

Ah yeah I missed that this can still be the case then. Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

[lforst]

Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

Considering this is an invariant that should never happen in the first place I'd rather have it show up and cause people to ask questions rather than hide it behind "wrong" data.

[lforst]

We're seeing a weird issue where error events have the wrong culprit containing incomplete-app-router-transaction.

My suspicion os that a transaction is started, then an error happens in between of the transaction being started and the transaction name being updated, and it thus receives the wrong culprit.

We fix this by not starting the transaction with a wrong name in the first place. I dunno why I didn't do this from the beginning.

I didn't fully verify this but it would be the only explanation I can come up with and I think it is nicer this way in any case.

Note that I even saw this happening with router.push transactions where we don't go into the branch where we don't update the name so it's not related to that code path.

[chargome]

Ok we need to stop working on the same issues in parallel 😂

What I found so far that this is mainly an issue on safari mobile (could e.g. be a failing preflight request) and we only filter on this transaction name for type "transaction"

So I guess we can also remove the event processor with this pr!

ref #15411

[lforst]

Oh whoops, I zoned out in the meeting and thought you were talking about not-found errors or something because I heard event processor and forgot that we had the one for transactions. My bad!

I wouldn't remove the transaction event processor because it serves a different purpose than this fix. This fix is about making sure errors receive the right culprit, the processor is about dropping transactions for dropping transactions that never receive the popevent call - for example when the router.pop is called but the actual navigation never completes and the popevent isn't fired.

[chargome]

Ah yeah I missed that this can still be the case then. Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

[lforst]

Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

Considering this is an invariant that should never happen in the first place I'd rather have it show up and cause people to ask questions rather than hide it behind "wrong" data.

[lforst]

We're seeing a weird issue where error events have the wrong culprit containing incomplete-app-router-transaction.

My suspicion os that a transaction is started, then an error happens in between of the transaction being started and the transaction name being updated, and it thus receives the wrong culprit.

We fix this by not starting the transaction with a wrong name in the first place. I dunno why I didn't do this from the beginning.

I didn't fully verify this but it would be the only explanation I can come up with and I think it is nicer this way in any case.

Note that I even saw this happening with router.push transactions where we don't go into the branch where we don't update the name so it's not related to that code path.

[chargome]

Ok we need to stop working on the same issues in parallel 😂

What I found so far that this is mainly an issue on safari mobile (could e.g. be a failing preflight request) and we only filter on this transaction name for type "transaction"

So I guess we can also remove the event processor with this pr!

ref #15411

[lforst]

Oh whoops, I zoned out in the meeting and thought you were talking about not-found errors or something because I heard event processor and forgot that we had the one for transactions. My bad!

I wouldn't remove the transaction event processor because it serves a different purpose than this fix. This fix is about making sure errors receive the right culprit, the processor is about dropping transactions for dropping transactions that never receive the popevent call - for example when the router.pop is called but the actual navigation never completes and the popevent isn't fired.

[chargome]

Ah yeah I missed that this can still be the case then. Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

[lforst]

Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

Considering this is an invariant that should never happen in the first place I'd rather have it show up and cause people to ask questions rather than hide it behind "wrong" data.

[lforst]

We're seeing a weird issue where error events have the wrong culprit containing incomplete-app-router-transaction.

My suspicion os that a transaction is started, then an error happens in between of the transaction being started and the transaction name being updated, and it thus receives the wrong culprit.

We fix this by not starting the transaction with a wrong name in the first place. I dunno why I didn't do this from the beginning.

I didn't fully verify this but it would be the only explanation I can come up with and I think it is nicer this way in any case.

Note that I even saw this happening with router.push transactions where we don't go into the branch where we don't update the name so it's not related to that code path.

[chargome]

Ok we need to stop working on the same issues in parallel 😂

What I found so far that this is mainly an issue on safari mobile (could e.g. be a failing preflight request) and we only filter on this transaction name for type "transaction"

So I guess we can also remove the event processor with this pr!

ref #15411

[lforst]

Oh whoops, I zoned out in the meeting and thought you were talking about not-found errors or something because I heard event processor and forgot that we had the one for transactions. My bad!

I wouldn't remove the transaction event processor because it serves a different purpose than this fix. This fix is about making sure errors receive the right culprit, the processor is about dropping transactions for dropping transactions that never receive the popevent call - for example when the router.pop is called but the actual navigation never completes and the popevent isn't fired.

[chargome]

Ah yeah I missed that this can still be the case then. Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

[lforst]

Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

Considering this is an invariant that should never happen in the first place I'd rather have it show up and cause people to ask questions rather than hide it behind "wrong" data.

[lforst]

We're seeing a weird issue where error events have the wrong culprit containing incomplete-app-router-transaction.

My suspicion os that a transaction is started, then an error happens in between of the transaction being started and the transaction name being updated, and it thus receives the wrong culprit.

We fix this by not starting the transaction with a wrong name in the first place. I dunno why I didn't do this from the beginning.

I didn't fully verify this but it would be the only explanation I can come up with and I think it is nicer this way in any case.

Note that I even saw this happening with router.push transactions where we don't go into the branch where we don't update the name so it's not related to that code path.

[chargome]

Ok we need to stop working on the same issues in parallel 😂

What I found so far that this is mainly an issue on safari mobile (could e.g. be a failing preflight request) and we only filter on this transaction name for type "transaction"

So I guess we can also remove the event processor with this pr!

ref #15411

[lforst]

Oh whoops, I zoned out in the meeting and thought you were talking about not-found errors or something because I heard event processor and forgot that we had the one for transactions. My bad!

I wouldn't remove the transaction event processor because it serves a different purpose than this fix. This fix is about making sure errors receive the right culprit, the processor is about dropping transactions for dropping transactions that never receive the popevent call - for example when the router.pop is called but the actual navigation never completes and the popevent isn't fired.

[chargome]

Ah yeah I missed that this can still be the case then. Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

[lforst]

Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

Considering this is an invariant that should never happen in the first place I'd rather have it show up and cause people to ask questions rather than hide it behind "wrong" data.

[lforst]

We're seeing a weird issue where error events have the wrong culprit containing incomplete-app-router-transaction.

My suspicion os that a transaction is started, then an error happens in between of the transaction being started and the transaction name being updated, and it thus receives the wrong culprit.

We fix this by not starting the transaction with a wrong name in the first place. I dunno why I didn't do this from the beginning.

I didn't fully verify this but it would be the only explanation I can come up with and I think it is nicer this way in any case.

Note that I even saw this happening with router.push transactions where we don't go into the branch where we don't update the name so it's not related to that code path.

[chargome]

Ok we need to stop working on the same issues in parallel 😂

What I found so far that this is mainly an issue on safari mobile (could e.g. be a failing preflight request) and we only filter on this transaction name for type "transaction"

So I guess we can also remove the event processor with this pr!

ref #15411

[lforst]

Oh whoops, I zoned out in the meeting and thought you were talking about not-found errors or something because I heard event processor and forgot that we had the one for transactions. My bad!

I wouldn't remove the transaction event processor because it serves a different purpose than this fix. This fix is about making sure errors receive the right culprit, the processor is about dropping transactions for dropping transactions that never receive the popevent call - for example when the router.pop is called but the actual navigation never completes and the popevent isn't fired.

[chargome]

Ah yeah I missed that this can still be the case then. Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

[lforst]

Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

Considering this is an invariant that should never happen in the first place I'd rather have it show up and cause people to ask questions rather than hide it behind "wrong" data.

[lforst]

We're seeing a weird issue where error events have the wrong culprit containing incomplete-app-router-transaction.

My suspicion os that a transaction is started, then an error happens in between of the transaction being started and the transaction name being updated, and it thus receives the wrong culprit.

We fix this by not starting the transaction with a wrong name in the first place. I dunno why I didn't do this from the beginning.

I didn't fully verify this but it would be the only explanation I can come up with and I think it is nicer this way in any case.

Note that I even saw this happening with router.push transactions where we don't go into the branch where we don't update the name so it's not related to that code path.

[chargome]

Ok we need to stop working on the same issues in parallel 😂

What I found so far that this is mainly an issue on safari mobile (could e.g. be a failing preflight request) and we only filter on this transaction name for type "transaction"

So I guess we can also remove the event processor with this pr!

ref #15411

[lforst]

Oh whoops, I zoned out in the meeting and thought you were talking about not-found errors or something because I heard event processor and forgot that we had the one for transactions. My bad!

I wouldn't remove the transaction event processor because it serves a different purpose than this fix. This fix is about making sure errors receive the right culprit, the processor is about dropping transactions for dropping transactions that never receive the popevent call - for example when the router.pop is called but the actual navigation never completes and the popevent isn't fired.

[chargome]

Ah yeah I missed that this can still be the case then. Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

[lforst]

Should we then adapt the processor to maybe fallback to the path whenever we face an error event with this tx name?

Considering this is an invariant that should never happen in the first place I'd rather have it show up and cause people to ask questions rather than hide it behind "wrong" data.