ping

Looks good to me, but I wonder whether

KWallet writes an error to stderr when a 0-winId is given to openWallet() which is received by git-lfs and aborts the operation.

should be fixed more in a more general way as well.

In D9072#189303, @fvogt wrote:

Looks good to me, but I wonder whether

KWallet writes an error to stderr when a 0-winId is given to openWallet() which is received by git-lfs and aborts the operation.

should be fixed more in a more general way as well.

Like how ?

In D9072#189550, @mkoller wrote:

In D9072#189303, @fvogt wrote:

Looks good to me, but I wonder whether

KWallet writes an error to stderr when a 0-winId is given to openWallet() which is received by git-lfs and aborts the operation.

should be fixed more in a more general way as well.

Like how ?

Duping stdout/stderr to new file descriptors, using those and closing 1/2 would be very effective.

Does kwallet really write to stderr? If so, why does git-lfs parse stderr in addition to stdout?

so you mean instead of a one line change fiddling with file descriptors is a better way ?
I don't think so.

And effectively I was told that passing a 0 pointer to kwallet is wrong.
https://mail.kde.org/pipermail/kde-frameworks-devel/2017-May/045085.html

• Does kwallet really write to stderr?

yes, using qDebug()

• If so, why does git-lfs parse stderr in addition to stdout?

you have to ask the developers of git-lfs. I don't know.