Re: [PATCH 0/5] transport-helper: serious crash fix
Felipe Contreras felipe.contre...@gmail.com writes: One of the most serious recurring issues[1][2][3] with remote helpers is that marks get out of sync. The way to analize and reproduce these wasn't trivial, but the culprit seems to be a crash while doing `git push`. It has been known already how exactly that happens, but no simple way how to fix it. This is the simplest way so far; tell `git fast-export` to export the marks to a temporary file, and move it to the right location only *after* the remote helper has finished its job without errors. Since the code wasn't prepared for a change like this, some reorganization changes are needed. More changes might be welcome to further propagate the errors properly through the code, but for the moment the errors are propagated to the right location, in order to fix this specific problem. This seems to be based on a somewhat older codebase; I tried to be careful while adjusting the patch to the current codebase, but please give it an eyeball to see if I didn't make any silly mistake in the version that has been parked on 'pu'. After an positive Ack (or you botched the forward-porting; here is a rebased one you should replace them with), I'd like to merge it to 'next' and then to 'master' by -rc0 (or -rc1 at the latest) so that we have enough time to expose the updated code before 2.0. Thanks. -- To unsubscribe from this list: send the line unsubscribe git in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
[PATCH 0/5] transport-helper: serious crash fix
One of the most serious recurring issues[1][2][3] with remote helpers is that marks get out of sync. The way to analize and reproduce these wasn't trivial, but the culprit seems to be a crash while doing `git push`. It has been known already how exactly that happens, but no simple way how to fix it. This is the simplest way so far; tell `git fast-export` to export the marks to a temporary file, and move it to the right location only *after* the remote helper has finished its job without errors. Since the code wasn't prepared for a change like this, some reorganization changes are needed. More changes might be welcome to further propagate the errors properly through the code, but for the moment the errors are propagated to the right location, in order to fix this specific problem. [1] http://article.gmane.org/gmane.comp.version-control.git/223962 [2] http://article.gmane.org/gmane.comp.version-control.git/241707 [3] https://github.com/felipec/git/issues/56 Felipe Contreras (5): transport-helper: remove barely used xchgline() remote-helpers: make recvline return an error transport-helper: propagate recvline() error pushing transport-helper: trivial cleanup transport-helper: fix sync issue on crashes t/t5801-remote-helpers.sh | 17 ++- transport-helper.c| 72 --- 2 files changed, 59 insertions(+), 30 deletions(-) -- 1.9.1+fc3.9.gc73078e -- To unsubscribe from this list: send the line unsubscribe git in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html