Bug 1539133 Comment 11 Edit History

Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.

So the summary is that:

a) We have a clownshoes issue because the code is allocating a 1M arraybuffer, which happens to 1) go through to our allocator unmodified because ArrayBuffers are special 2) hits the worst case there.

b) The code is allocating 35GB. Due to the clownshoes, this doubles to 70GB, which is over the 654 limit that Linux defaults to.

The clownshoes are unfortunate but I don't see a simple (or complicated) way around it. Unless there's evidence that this worsens performance in the field, I think this is a WONTFIX from my side. Workaround would be to make the buffer 4096 bytes smaller.

If you need to keep the testcase unmodified, I guess you'll want to look if the kernel settings on CI machines can be modified.
So the summary is that:

a) We have a clownshoes issue because the code is allocating a 1M arraybuffer, which happens to 1) go through to our allocator unmodified because ArrayBuffers are special 2) hits the worst case there.

b) The code is allocating 35GB. Due to the clownshoes, this doubles to 70GB, which is over the 65G limit that Linux defaults to.

The clownshoes are unfortunate but I don't see a simple (or complicated) way around it. Unless there's evidence that this worsens performance in the field, I think this is a WONTFIX from my side. Workaround would be to make the buffer 4096 bytes smaller.

If you need to keep the testcase unmodified, I guess you'll want to look if the kernel settings on CI machines can be modified.
So the summary is that:

a) We have a clownshoes issue because the code is allocating a 1M arraybuffer, which happens to 1) go through to our allocator unmodified because ArrayBuffers are special 2) hits the worst case there.

b) The code is allocating 35GB. Due to the clownshoes, this doubles to 70GB, which is over the 65G limit that Linux defaults to.

The clownshoes are unfortunate but I don't see a simple (or complicated) way around it. Unless there's evidence that this worsens performance in the field, I think this is a WONTFIX from my side. Workaround would be to make the buffer 4096 bytes smaller or allocate a bit less in a single run.

If you need to keep the testcase unmodified, I guess you'll want to look if the kernel settings on CI machines can be modified.

Back to Bug 1539133 Comment 11