adrianschmutzler:

+1

This is completely independent from Luci, and should be addressed directly in uci project most probably.

richb-hanover:

Hi Adrian,

Thanks for thinking about this.

I'm confused, though. I filed a feature request at openwrt/luci#3569 but was told (I think) to file the request over here.

Where should this request go? Many thanks.

Rich

adrianschmutzler:

It's okay here, as the uci system is part of OpenWrt.

richb-hanover:

That's great. Thank you.

adrianschmutzler:

Those two lines are relevant.

https://git.openwrt.org/?p=project/uci.git;a=blob;f=libuci.c;h=140edf2adbb0817ad89fdbb2d78984290d3017ca;hb=HEAD#l33
https://git.openwrt.org/?p=project/uci.git;a=blob;f=libuci.c;h=140edf2adbb0817ad89fdbb2d78984290d3017ca;hb=HEAD#l218

I will have a look whether I can manage to write pure C code ...

adrianschmutzler:

Will test this when I have time on device:
diff --git a/libuci.c b/libuci.c
index 140edf2..3099f20 100644
--- a/libuci.c
+++ b/libuci.c
@@ -210,12 +210,16 @@ int uci_load(struct uci_context *ctx, const char *name, struct uci_package **pac
int uci_set_backend(struct uci_context *ctx, const char *name)
{
struct uci_element *e;

}


adrianschmutzler:

Looks like it's not that easy.

adrianschmutzler:

Seems like it's more complicated:
One would have to write the name of the missing file to error_info in uci_get_errorstr(), like done fore UCI_ERR_PARSE already:
https://git.openwrt.org/?p=project/uci.git;a=blob;f=libuci.c;h=140edf2adbb0817ad89fdbb2d78984290d3017ca;hb=HEAD#l141

The problem is that the information about the requested data is not in the uci_context struct (ctx), but in an ubi_ptr struct. Definitions here:
https://git.openwrt.org/?p=project/uci.git;a=blob;f=uci.h;h=43f2052a00e6901e67cd6d1956fdb36375b41358;hb=HEAD#l396
https://git.openwrt.org/?p=project/uci.git;a=blob;f=uci.h;h=43f2052a00e6901e67cd6d1956fdb36375b41358;hb=HEAD#l396

So, one would have to create a field for the failed ptr or package string, and then write content to it for each situation where UCI_ERR_NOTFOUND is triggered. That seems to be what's done for the parsing error right now ...

This looks to me like it will require a lot of testing, and I do not have the time to implement it ATM.