Restoring files from a corrupt bundle

Erik Muttersbach erik at muttersbach.net
Mon Nov 29 15:07:22 CST 2010


I just found out that the size of the file I commted was 3050086400 Bytes, so about 2.8GB. You wrote that you use 31bit, so this probably caused an overflow?

Erik

On Nov 29, 2010, at 9:27 PM, Matt Mackall wrote:

> On Mon, 2010-11-29 at 21:17 +0100, Erik Muttersbach wrote:
>> Thanks for the link. I uncompressed the bundle and have a more or less
>> readable file.
> 
> How big is the result? The individual deltas in the bundle won't be
> compressed, so if it's not 10G, it won't contain your large files.
> 
>> Some parts seem to me to be binary:
>> 
>> H<8b>E<90>E1íL^A
>> I<8d><96>^H+^@^@H<89>×H<89><95>xÿÿÿè^@^@^@^@I<8b><8e>è)^@^@H<8b><95>xÿÿÿH<83>éhë^Oë^B<90><90>ÇA`^@^@^@^@H<8d>O<98>H<8b>yhL<8d>AhM9Ç^O^X^OuåH<89>ÆH<89>×M<89>¾è)^@^@M<89>¾ð)^@^@è^@^@^@^@L<89>÷ètìÿÿH<89>ßè^@^@^@^@IcÅE<85>íI^ODÄH<83>Äh[A\A]A^A_ÉÃë^A<90>A<8b>¶^@*^@^@º^A^@^@^@L<89>÷èüêÿÿ<85>À^O<84>«^B^@^@I<8d><86>Ø)^@^@I;<86>Ø)^@^@AÇ<86>^D*^@^@^A^@^@^@AÇ<86>^H*^@^@^B^@^@^@^O<85>Òþÿÿë(ë^F<90><90><90><90><90><90>H<89>ßè^@^@^@^@H<83>Äh1À[A\A]A^A_ÉÃë^E<90><90><90><90><90>I<8d><96>è)^@^@I;<96>è)^@^@tDI<89><86>Ø)^@^@I<89><86>à)^@^@1ÀI<89><96>è)^@^@I<89><96>ð)^@^@f<90>AÇ<84>^F8^Z^@^@^@^@^@^@AÇ<84>^Fè^Y^@^@^@^@^@^@H<83>è<80>H=^@^P^@^@uÜL<89>÷è<84>ëÿÿéGþÿÿë^E<90><90><90><90><90>AöE9^Hë^A<90>^O<85>^Z^B^@^@A^O·¾<9a>*^@^@iÿè^C^@^@è^@^@^@^@M;¾è)^@^@^O<84>â^@^@^@H<85>À^O<88>Ö^A^@^@A<8b><96>
>> *^@^@öÂ^BujH<85>Àt
> 
> Those look like our binary delta chunks.
> 
> Much to my surprise, it is possible to commit files > 2G and even verify
> and clone the result locally. And you can even commit changes to such
> files, but you won't be able to check out those changes as the delta
> applying code will get confused. I'll be adding a check to make sure we
> refuse to commit such large files in the future.
> 
> -- 
> Mathematics is the supreme nostalgia of our time.
> 
> 



More information about the Mercurial mailing list