Error on server while publishing artifacts should result in the build failure, so this issue should concern agent.

We probably need an option with textarea to specify mandatory artifacts. If any of the artifacts is missing, the build should fail with appropriate failure message.
If a wildcard is specified, any file that satisfies the wildcard should satisfy the condition.
Multiple values in the area should be treated as "AND conditions.


Alternative (and more obscure) approach is to introduce some special pattern into the artifacts publishing value to treat some artifacts as mandatory.

Or just have an option "Fail the build if no artifacts are produced".

I too have seen this problem - a build marked as successful when it failed to publish artifacts. (Not obvious why not - the corresponding build log is below). Naturally, all the tasks with artifact dependencies then failed.

[21:05:00]: [Publishing artifacts] Paths to publish: [src/eqsval-dist/**/*, teamcity-info.xml]
[21:05:00]: [Publishing artifacts] Publishing artifacts 'src/eqsval-dist/**/*' to root artifacts directory
[21:05:35]: [Publishing artifacts] Failed to publish files: Failed to publish artifacts. Server status: 500 (Internal Server Error)

Most probably it's caused by exceeding the maximum artifact size. The log should be more verbose about it, we were struggling with this problem for a couple of hours.
Failing the build is the right way to go with this issue.
Tested on 4.5.5

One more related feature request about fail build if there is no artifacts produced: http://www.jetbrains.net/devnet/message/5252766

Also requested in http://www.jetbrains.net/devnet/message/5260050

This happened on our server today. The artifact failed to publish but the build is marked as successful causing other dependencies to start and failed.

[Publishing artifacts] Failed to publish files: Failed to publish artifacts. Server status: 500 (Internal Server Error)

The artifact is 27mb in size and we have unlimited artifact size setting. So it's not the size issue. But anyway, we should have the option to fail the build if artifact publishing fails.

Because wild cards are allowed to define artifacts, the build could be set to fail on:

• number of artifacts,
  
• minimum total artifacts size (already collected for stats).
  

Current workaround to fail a build that did not produce any artifacts is to include the checking loging inside the build script right before the build finish and issue change the build status from within the build.

Also requested in http://devnet.jetbrains.net/message/5277847

The las couple of comments seem to misunderstand the issue - the problem isn't failing builds where artifacts have not been produced (the build script can, and should, do that). The problem is failing builds where the artifacts have been produced, but publishing them fails.

Guys,

Sorry for some mess up with the related issues, this issue actually got commented on two problems.

Let's make this issue address feature to fail build if it did not produce artifacts for some reason.
And TW-12243 will then handle any explicit errors that occurred during uploading artifacts to the server.

It seems perverse to take a long standing issue about failing to publish and convert it to one about failing to produce artifacts. And TW-12243 seems to be about a specific cause of failing to publish - would it not be better to mark that as a duplicate?

I've voted for this issue thinking it about publishing - I don't want that to count as a vote about not producing artefacts.

Alan,

This issue was marked as a feature and had a comment on the current issue's sense since November 2008.
Since then it got comments (and may be votes) for both the current sense and TW-12243.

If you are not interested in the current sense and care more about TW-12243, please unvote this one and vote TW-12243.
Sorry for the inconvenience, if any.

Yegor, I'll repeat TW-12243 seems to be about a specific cause of failing to publish (artefacts too big) I've not encountered that case and don't care to vote for it. As Pavel Sher says in the comments there this item covers the "failure to publish cases better" (at least it used to) and votes have been directed here instead.

I don't know why I got "Server status: 500 (Internal Server Error)" but it had nothing to do with artefact size (or space on server) as that didn't change between earlier or later builds that published successfully.

Could you at least replace the description of TW-12243 with the original text from this one? At least then it would cover the case I have trouble with.

Is there a update on this one being taken care about - 'Failing the build when all the artifacts are not published'

I've added TW-15084 to describe the problem orignally posted here

also requested in http://devnet.jetbrains.net/thread/292927

Currently, it is possible to fail a build if the combined size of build's artifacts is less/greater then specified. This is only partly what was described in this issue.
See also the comment.