[pacman-dev] [PATCH 2/2] ensure package download sizes are invalidated when an error occurs

[Xavier Chantry]

On Sun, May 16, 2010 at 10:22 PM, Jonathan Conder <j at skurvy.no-ip.org> wrote:
> Hi
>
> On Sun, 2010-05-16 at 14:54 +0200, Nagy Gabor wrote:
>> The spirit of this patch is better, but it is still not perfect, because
>> of the -Sw (PM_TRANS_FLAG_DOWNLOADONLY) case. That return(0) can also be
>> changed to goto error, and I can see one more RET_ERR in the function.
>
> See my other patch for the -Sw case. I seem to have a tendency to
> overlook the RET_ERR ones though...
>
>> I just note that in case of error-free -S commit operation, this
>> code-part is needless/overkill, because the target pkgcache packages are
>> replaced in the integrity-check part by the pmpkg_t of pkg.tar.xz file
>> (line 949), and then every download_size is either 0 or unset.
>
> Ok, good point, I didn't notice that. In that case I think we should
> move this back to where it was originally, but make sure only packages
> for the current repo get unset. This might be a good time to introduce a
> pmpkg_t parameter to the download callback as well.
>
>> An other side note:
>> I have some bad feelings when I see many foo->download_size in this
>> codepart. We have a good access function for this. I know that atm all
>> download_sizes are filled in correctly at the end of sync_prepare, but
>> that may be changed later. But this direct access causes some nanosec
>> speedup, so I dunno. :-)
>
> I agree, but I don't mind either way. Since PM_ERR_PKG_INVALID_NAME is
> checked in the commit I'm not sure we even need the pre-computation any
> more. I'll have a closer look when I get the chance.
>

I did not really follow the discussion, and still don't have much
time, sorry for that.

I just wonder if this is related to the following debug message I saw
with scriptlet002 :

debug: returning error 32 from compute_download_size : package
filename is not valid
Remove (1): dummy-1.0-1

Total Removed Size:   0.00 MB

Do you want to remove these packages? [Y/n]
removing dummy...
debug: removing package dummy-1.0-1