"Dumping very large path" message isn't helpful

[copumpkin]

It would be helpful to know which path is being dumped and ideally what led to the dumpage. Given where the warning is emitted right now in the code, it's not obvious to me how to add that detail to the message, but it's pretty unfriendly right now. I guess we could add a constructor parameter to an FdSink that tells it some human-meaningful representation of the thing being written to it?

cc @edolstra @shlevy

[shlevy]

Yeah, a message in FdSink constructor seems the way to go.

[shlevy]

(well, actually, using constant space for dumping is obviously the way to go 😛 )

[copumpkin]

#619

[domenkozar]

@edolstra how come 01d33ea never landed in master?

[edolstra]

@domenkozar That patch doesn't apply cleanly without cf4a7ae.

[domenkozar]

Having both would be really great :)

[domenkozar]

@edolstra would you mind explaining what withholds these two commits back?

These are two of the most annoying bugs I encounter when using Nix and even being experienced they are annoying to overcome.

I think it's better to make it work&correct and we can correct the performance one day if needed.

[zimbatm]

This is still an issue today. I just got that error message and don't know what path is causing the issue.

[domenkozar]

Same here today, nix-daemon ate all my free of memory and I have no idea what path caused it.

[stale]

I marked this as stale due to inactivity. → More info

[Ericson2314]

I think the prereq patch has landed on master in some form? This might also be straight up fixed on master too.

[aleb]

cf4a7ae landed as 34f2512

Any chance it can be backported to the 2.3-maintenance branch used for making the 2.3.* releases https://github.com/NixOS/nix/blob/2.3-maintenance/src/libstore/local-store.cc ?

The #2981 can be closed as it's forgotten and this different fix has been merged instead.

[stale]

I marked this as stale due to inactivity. → More info

[pingiun]

this isn't stale, as the message is still shown and still not helpful