Reopened #1471.

This silently reloads files that have unsaved changes. The changes can then be recovered by invoking “undo” in each affected document, but I think this is dangerous. I could invoke “reload all” by accident and then I don’t even know what to undo.

Maybe a dialog box like this would be appropriate:

The following 3 files have unsaved changes:

/home/vasiliy/foo/bar/baz /home/vasiliy/quux/xyzzy /home/vasiliy/wololo

Do you want to reload them as well? The changes will be gone, but you can restore them by invoking “undo” in individual documents.

<Yes, reload these files>     <No, skip these files>

Personally, I would also be OK with silently skipping such files — that’s what I do now in my plugin. But at least a message in the status bar would be useful and easy to add.

After all documents are reloaded, I end up in a different document than the one I invoked “reload all” in. I think you should remember the initial document and then switch back to it.

This may be out of scope for your PR, but personally, for [my use cases](https://github.com/geany/geany/pull/1246#issuecomment-298176055), the ideal feature would be “reload all documents that have changed on disk” — to avoid wasting time on reloading documents that have not actually changed.

@attila-v pushed 2 commits.

1b3918f Fix 'Reload all' texts 75e43f8 Use 'foreach_document()' definition instead of using own iteration cycle

@vfaronov I have made the fixations you asked before. I examined the code of geany, and I think it's complicated to solve reloading the changed files only. When geany recognizes, that a file has different time mark on the disk than in a tag of struct of document, it immediately write back the new time into struct (document.c, row 3732). So I could not understood the difference when the reloading process is running. The solution is probably storing this information (diff is true/false) in a new tag in the struct, or checking again the time mark of the actual file in the time of reloading process. The first needs more memory allocations, the second needs more runtime. Sorry for my english, I hope it was clear for you what I wanted to tell you. cheers, Attila

@attila-v pushed 1 commit.

0963dbb Check file status before reload it

@elextr thank you for your suggestions. I had a little time, and (I think) I ended and fixed this feature. I put a new button to the reload message to obtain well working feature. I hope you enjoy it.

@attila-v Thank you for your work! I will try running Geany with this change to see how well it works for me.

So far I see a couple problems.

The first problem is that you still silently lose unsaved changes (in files that have diverged from disk).

The second problem is an assertion error. I can reproduce it as follows:

1. Open some file in Geany. 2. Delete it from disk. 3. Switch to Geany. At the top of the document, Geany shows an orange info bar that says: “File ... was not found on disk! Try to resave the file?”

With your change, when this orange info bar appears, I see the following in the console:

GtkWarning: IA__gtk_info_bar_add_button: assertion 'button_text != NULL' failed

Without your change, there is no such warning.

@attila-v As I said before, I think this can be a problem even when `keep_edit_history_on_reload` is enabled. Normal reload works per-document, so I know exactly which files have been reloaded, so I can go and undo there. But with reload-all, how do I know where to undo?

Perhaps a minimal fix would be to note such files (reloaded despite unsaved changes) in the status bar, so I could see them all in the “Status” pane.

@attila-v pushed 1 commit.

1e11eb0 Move button initializer into condition

@vfaronov I solved the gtk assertion. The problem was, that I fitfully made buttons from original version. Thank you for your detection. And I also put a message into status bar when unsaved file has reloaded. I hope you thought so.

I have been running with this change for some time. So far it seems to work well.

(Just reporting my experience; the PR still needs a review by maintainers.)

It is working like the simple reload function: if the focus is on the unsaved document, then both functions reload it (and send message to the status); but if the focus is on the other document, then does not happen anything with unsaved document. When unsaved document reloaded, you can undo it to the unsaved status of course (if the 'keep_edit_history_on_reload' box is checked in the Preferences).