[PATCH] Check the list of enabled mods when running a replay from the command line

[Itms]

Right now the replay mode fails when no mod is specified, we should check that and inform the user about what is happening.

Maybe this is part of a more general task, which is including mod data in commands.txt and other logging stuff.

We can use the list of mods in the commands.txt to make sure everything is fine.

[mimo]

It works for me on linux.

[Itms]

Apparently it's the VFS not finding the file, maybe related to mod selection changes?

I update the ticket description.

[Itms]

Um, I just understood that now the replay mode needs to know which mod(s) to load. I'm updating the docs and this ticket.

[elexis]

Trivial, just show the user a hint if g_args has "replay" but no "mod" argument. File: main.cpp

[Arno Hemelhof]

Displays an error message when you try to view a replay without specifying any mods

[Arno Hemelhof]

This patch just implements an error message. Should something else be done when nothing is specified, like loading public?

[elexis]

• Looks like it works, but in fact its broken as the XML library needs to be deinitialized on shutdown, see that CXeromyces::Terminate(); call below.
• should use LOGERROR (without newline at the end)
• "None were detected" doesn't add any information that isn't contained in the previous sentence.
• The message is clearly directed at users who are not familiar with the code or technical background. So the term "official mod" is a bit confusing / counterintuitive and should be avoided. I recall seeing "Did you mean this?" being printed out in some tool.

At least one mod should be specified! None were detected!\nIf you intended to use the official mod, do so by adding the argument '-mod=public'

• Maybe it would be better to first load the replay, extract the mods that specific replay wants, attempt to load them and then throw an error message if the specified mods were not found. Trying to replay something with a different set of mods doesn't seem to make any sense and if some developer wants to hack something with mods, he can just hack the replay file. This would be true for the visual replay as well.

[Arno Hemelhof]

Changed based on the remarks of comment 7

[Arno Hemelhof]

Thank's for the review. I based myself on the way it is handled when the replay file isn't found (Line 443) and CXeromyces::Terminate() isn't being called there. I have added the terminate call, used LOGERROR and rephrased my error message.

[Itms]

Hi, and thanks for your work so far. :)

When this ticket was created, we didn't have the list of enabled mods included in the commands.txt file. Now that we have that, it should be possible to check that the list of enabled mods is the same as in the commands file.

Currently the check is done in the visual replay menu but not when using the command line. I think a mismatch in the mod lists should trigger a warning instead of an error, so people using the command line are still allowed to do what they want.

This would be a more complex task. Would you be up for it? You can come to our IRC channel if you need more information.

Also, for next time, please take a look at our new way of SubmittingPatches: we're making changes to our review process.

[Arno Hemelhof]

I have written a new patch. The plan was to automatically load the mods specified in the mod list of the replay file, but to parse the JSON, I needed a JSContext*.The context of the Simulation2 instance however, is created in the constructor of CGame which requires the mods to already be loaded. To change the order in which these action occur, a rewrite would be needed. Since this is my first ticket, I don't think I have enough experience to do such a rewrite. A new ticket could be made for this feature if necessary.

This patch prints out an error message when the replay is trying to be loaded without any mods specified. It also displays a warning which shows which mods are missing based on the mod list in the replay file,and which mods were loaded which aren't in the replay file. Would it be better to let the program continue when no mods are loaded so it can print out which mods are missing? The program would still crash in that case.

[elexis]

Sounds right, I'll take a look.

[elexis]

One more thing though. There is this mock mod "user" which is launched in the release version but not in the svn working copy. See the logic of hasSameMods in functions_utility_loadsave.js. The mod should just be ignored, it is not relevant in this context.

[Arno Hemelhof]

Now it ignores the mod "user" in both loops.

[Arno Hemelhof]

A new version which resolves a file confict (to make applying and reviewing easier)

[elexis]

In 18574:

Print warnings when no or the wrong list of mods is passed when starting non-visual replays. Patch by ArnH, fixes #2807.

[elexis]

Thanks for the patch! I have removed the whitespace in emptylines and unneeded comments.