Welcome to the Splat alpha!

The string rf9e2887948e8a1632f07ab179945027884e1f6f4^1 does not look like a valid SHA. Could you show us the relevant diff lines for the .hftags file?

• Status:

  +	NeedInfo

I added another review with same problem.

The file '.hgtags' (r2b2ebdd19c379afb4cf48618ee4f319d93c7e7cf^1) could not be found in the repository

Patch is attached.

• Files:

  +	rb367.patch

Got this in logfile:

Reviewboard_1 | 2015-07-16 19:44:18,074 - ERROR -  - 404
Reviewboard_1 | 2015-07-16 19:44:18,759 - ERROR -  - HTTP error code 400 when fetching file from http://scmmanager/scm/hg/REPO/raw/2b2ebdd19c379afb4cf48618ee4f319d93c7e7cf^1/.hgtags: HTTP Error 400: Bad Request
Reviewboard_1 | 2015-07-16 19:44:19,461 - ERROR -  - HTTP error code 400 when fetching file from http://scmmanager/scm/hg/REPO/hg-history/2b2ebdd19c379afb4cf48618ee4f319d93c7e7cf^1/.hgtags: HTTP Error 400: Bad Request

Ping! Do you need some more info?

Sorry I haven't had time to look at this bug.

What are you using to serve your repo, hg web?

We use https://www.scm-manager.org for this. It will provide a hgweb, yes.

SCM Manager is a hosting service that Review Board currently does not support.

• Status:

  -	NeedInfo
  +	Confirmed

• Tags:

  -	Type-Defect
  +	Type-Enhancement

• Summary:

  +	Add support for SCM Manager as a HostingService

I dont' think that SCM-Manager is the problem! It spawns a python shell with hgweb of Mercurial.

Nevertheless... I tried "hg serve" with the repository and Reviewboard 2.0.18. It is still broken.

Access Log of "hg serve":
10.21.253.22 - - [14/Aug/2015 12:52:43] "GET /raw-file/f9e2887948e8a1632f07ab179945027884e1f6f4^1/.hgtags HTTP/1.1" 404 -
10.21.253.22 - - [14/Aug/2015 12:52:44] "GET /raw/f9e2887948e8a1632f07ab179945027884e1f6f4^1/.hgtags HTTP/1.1" 400 -
10.21.253.22 - - [14/Aug/2015 12:52:44] "GET /hg-history/f9e2887948e8a1632f07ab179945027884e1f6f4^1/.hgtags HTTP/1.1" 400 -

The "^1" is the problem. That's not a valid HG revision, when it comes to fetching files remotely. 

The diff file you posted doesn't match the revisions being referred to. Can you show an up-to-date diff with the accompanying errors, along with the command line used to generate the diff, and any steps you used to post it?

• Status:

  -	Confirmed
  +	NeedInfo

• Tags:

  +	SCM-Mercurial

Ok, I found the problem.

I'm using a script on server side [1] to create the diff, parent diff and the base commit id using MercurialClient class from rbtools 0.7.4 (tried 5985bfc7145fc240c9e23a99d80f1e149bec355b, too).

So the reference is a NODEID^1 that will produce correct diff and parent diff (without ^1) but MercurialClient returns the NODEID^1 as base_commit_id.


[1] https://reviews.reviewboard.org/r/7100/

Work-around:

diff --git a/rbtools/clients/mercurial.py b/rbtools/clients/mercurial.py
index ad0d1f9..2f6edc9 100644
--- a/rbtools/clients/mercurial.py
+++ b/rbtools/clients/mercurial.py
@@ -412,6 +412,9 @@ class MercurialClient(SCMClient):
             base_commit_id = revisions['base']
             parent_diff = None
 
+        base_commit_id = self._execute(
+             ['hg', 'log', '-r', base_commit_id, '-T {node}'], env=self._hg_env, results_unicode=False)
+
         return {
             'diff': diff,
             'parent_diff': parent_diff,

This introduces it: https://github.com/reviewboard/reviewboard/commit/6b34af11397d92ae5c396ed911840ae6b0392cfd

If in table "diffviewer_diffset" in column "base_commit_id" is already a NODEID^1 it is, of course, still broken for old reviews.

update diffviewer_diffset set base_commit_id=NULL where base_commit_id like '%^1';

How exactly are you generating the diff with MercurialClient? You may be generating it in an incorrect way. If you generate the diff through rbt, does it work as intended?

You can see that in "rbtools/hooks/mercurial.py" with "class MercurialDiffer" at that review request [1] from above.

It calls MercurialClient with "NODEID^1" as it is allowed on local hg clients like with local git clients.

hg log -r tip^1
hg log -r tip^2
hg log -r tip^3

git log -r HEAD~1
git log -r HEAD~2
git log -r HEAD~3

[1] https://reviews.reviewboard.org/r/7100/

This issue should be renamed since it's a bug in rbtools and not a feature request. :-)

Please add SCM-manager as a Hosting service for local scm-manager repositories!

This issue is not to add SCM-Manager support!! It's just an incorrect name!

A patch is under review at https://reviews.reviewboard.org/r/7823/

• Status:

  -	NeedInfo
  +	PendingReview

• Summary:

  -	Add support for SCM Manager as a HostingService
  +	Resolve HG Commit IDs before posting

Fixed in release-0.7.x (d66f628). This will ship in 0.7.6. Thanks!

• Status:

  -	PendingReview
  +	Fixed