• Steven Schveighoffer (17/17) Oct 15 2021 Write a pretty trivial program to test if maybe a regression happened on...
• Imperatorn (4/7) Oct 15 2021 But, wouldn't it be best if it just worked as intended. Caching
• Vladimir Panteleev (5/8) Oct 16 2021 The functionality is based on
• Imperatorn (4/7) Oct 16 2021 Why would decreasing the number of versions work? To little
• Basile B. (7/24) Oct 16 2021 4. "all versions" have for effect to automatically create a PR on

Steven Schveighoffer <schveiguy gmail.com> writes:

Write a pretty trivial program to test if maybe a regression happened on 
run.dlang.io, and it now always has a server error.

I believe the error is from a timeout because we have so many versions 
of the compiler in the list.

A couple ways to fix this:

1. Start incrementing the oldest version. Not ideal, but at least better 
than not working. Probably the easiest fix.
2. Most likely it got through most of the versions. So you can cache the 
partial result, and continue it on a subsequent run (it already caches 
completed results).
3. Change the dropdown to range the "previous versions" when they start 
timing out, i.e. instead of 2.060 - latest, you have 2.060 - 2.065, and 
2.066 - latest as a separate option.

I use this facility all the time to see if something is a regression or 
not, and it's no longer working. This is a wonderful mechanism that I 
definitely miss.

-Steve

Imperatorn <johan_forsberg_86 hotmail.com> writes:

On Friday, 15 October 2021 at 14:15:09 UTC, Steven Schveighoffer 
wrote:
 Write a pretty trivial program to test if maybe a regression 
 happened on run.dlang.io, and it now always has a server error.

 [...]

But, wouldn't it be best if it just worked as intended. Caching 
could be useful tho

Vladimir Panteleev <thecybershadow.lists gmail.com> writes:

On Friday, 15 October 2021 at 14:15:09 UTC, Steven Schveighoffer 
wrote:
 I use this facility all the time to see if something is a 
 regression or not, and it's no longer working. This is a 
 wonderful mechanism that I definitely miss.

The functionality is based on 
https://github.com/CyberShadow/misc/blob/master/dreg.d, an 
alternative would be to run the tool locally.

Imperatorn <johan_forsberg_86 hotmail.com> writes:

On Friday, 15 October 2021 at 14:15:09 UTC, Steven Schveighoffer 
wrote:
 Write a pretty trivial program to test if maybe a regression 
 happened on run.dlang.io, and it now always has a server error.

 [...]

Why would decreasing the number of versions work? To little 
memory / cpu / disk etc? 🤔

Basile B. <b2.temp gmx.com> writes:

On Friday, 15 October 2021 at 14:15:09 UTC, Steven Schveighoffer 
wrote:
 Write a pretty trivial program to test if maybe a regression 
 happened on run.dlang.io, and it now always has a server error.

 I believe the error is from a timeout because we have so many 
 versions of the compiler in the list.

 A couple ways to fix this:

 1. Start incrementing the oldest version. Not ideal, but at 
 least better than not working. Probably the easiest fix.
 2. Most likely it got through most of the versions. So you can 
 cache the partial result, and continue it on a subsequent run 
 (it already caches completed results).
 3. Change the dropdown to range the "previous versions" when 
 they start timing out, i.e. instead of 2.060 - latest, you have 
 2.060 - 2.065, and 2.066 - latest as a separate option.

 I use this facility all the time to see if something is a 
 regression or not, and it's no longer working. This is a 
 wonderful mechanism that I definitely miss.

 -Steve

4. "all versions" have for effect to automatically create a PR on 
a repo that run, as CI, the editor content on all the compilers, 
e.g using a special docker image.
when the user click "run" he's given a link to the (future) CI 
results.